| { | |
| "paper_id": "1993", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T07:35:52.467806Z" | |
| }, | |
| "title": "Integration of Morphological and Syntactic Analysis Based on LR Parsing Algorithm", | |
| "authors": [ | |
| { | |
| "first": "Tanaka", | |
| "middle": [], | |
| "last": "Hozumi", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Tokyo Institute of Technology", | |
| "location": { | |
| "addrLine": "2-12-1 Ookayama Meguro", | |
| "postCode": "152", | |
| "settlement": "Tokyo", | |
| "country": "Japan" | |
| } | |
| }, | |
| "email": "" | |
| }, | |
| { | |
| "first": "Tokunaga", | |
| "middle": [], | |
| "last": "Takenobu", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Tokyo Institute of Technology", | |
| "location": { | |
| "addrLine": "2-12-1 Ookayama Meguro", | |
| "postCode": "152", | |
| "settlement": "Tokyo", | |
| "country": "Japan" | |
| } | |
| }, | |
| "email": "" | |
| }, | |
| { | |
| "first": "Aizawa", | |
| "middle": [], | |
| "last": "Michio", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Tokyo Institute of Technology", | |
| "location": { | |
| "addrLine": "2-12-1 Ookayama Meguro", | |
| "postCode": "152", | |
| "settlement": "Tokyo", | |
| "country": "Japan" | |
| } | |
| }, | |
| "email": "" | |
| } | |
| ], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "Morphological analysis of Japanese is very different from that of English, because no spaces are placed between words. The analysis includes segmentation of words. However, ambiguities in segmentation is not always resolved only with morphological information. This paper proposes a method to integrate the morphological and syntactic analysis based on LR parsing algorithm. An LR table derived from grammar rules is modified on the basis of connectabilities between two adjacent words. The modified LR table reflects both the morphological and syntactic constraints. Using the LR table, efficient morphological and syntactic analysis is available.", | |
| "pdf_parse": { | |
| "paper_id": "1993", | |
| "_pdf_hash": "", | |
| "abstract": [ | |
| { | |
| "text": "Morphological analysis of Japanese is very different from that of English, because no spaces are placed between words. The analysis includes segmentation of words. However, ambiguities in segmentation is not always resolved only with morphological information. This paper proposes a method to integrate the morphological and syntactic analysis based on LR parsing algorithm. An LR table derived from grammar rules is modified on the basis of connectabilities between two adjacent words. The modified LR table reflects both the morphological and syntactic constraints. Using the LR table, efficient morphological and syntactic analysis is available.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Abstract", | |
| "sec_num": null | |
| } | |
| ], | |
| "body_text": [ | |
| { | |
| "text": "Cascade: Separate the morphological and syn-", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Morphological analysis of Japanese is very differ ent from that of English, because no spaces are placed between words. This is also the case in many Asian languages such as Korean, Chinese, Thai and so forth. In the Indo-European family, some languages such as German have the same phenomena in forming complex noun phrases . Processing such languages requires the identifica tion of the boundaries of words in the first place. This process is often called segmentation which is one of the most important tasks of morphological analysis for these languages . Segmentation is a very important process, since the wrong segmentation causes fatal errors in the later stages such as syntactic, semantic and contextual analysis. However, correct segmenta tion is not always possible only with morphologi cal information. Syntactic, semantic and contex tual information may help resolve the ambiguities in segmentation. Over the past few decades a number of studies have been made on the morphological and syntac tic analysis of Japanese . They can be classified into the following three approaches:", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "The morphological and syntactic constraints are represented sepa rately.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "tactic analysis and execute them in a cas cade manner.", | |
| "sec_num": "101" | |
| }, | |
| { | |
| "text": "Interleave: Separate the morphological and syntactic analysis and execute them inter leavingly. The morphological and syntactic constraints are represented separately.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "tactic analysis and execute them in a cas cade manner.", | |
| "sec_num": "101" | |
| }, | |
| { | |
| "text": "Single Framework: Represent both the mor phological and syntactic constraints in a single framework such as context free gram mars ( CFGs) and make no distinction be tween the two analysis . (Mine et al., 1990) . Most other systems use a connection matrix in stead of a regular grammar (Miyazaki et al., 1984; Sugimura et al., 1989) . The main drawbacks of these approaches are as follows:", | |
| "cite_spans": [ | |
| { | |
| "start": 191, | |
| "end": 210, | |
| "text": "(Mine et al., 1990)", | |
| "ref_id": "BIBREF3" | |
| }, | |
| { | |
| "start": 286, | |
| "end": 309, | |
| "text": "(Miyazaki et al., 1984;", | |
| "ref_id": "BIBREF4" | |
| }, | |
| { | |
| "start": 310, | |
| "end": 332, | |
| "text": "Sugimura et al., 1989)", | |
| "ref_id": "BIBREF7" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "tactic analysis and execute them in a cas cade manner.", | |
| "sec_num": "101" | |
| }, | |
| { | |
| "text": "\u2022 It may require two different algorithms for each analysis.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Representing the morphological and syntacti cal constraints separately as in the first two ap proaches, Cascade and Interleave, makes main taining and extending the constraints easier. This is an advantage of these approaches . Many natu ral language processing systems have used these two approaches. For example, Mine proposed a method to represent the morphological con straints in regular grammar and the syntactic constraints in CFG, and interleave the morpho logical and syntactic analysis", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u2022 It must retain all ambiguities from the mor phological analysis until the syntactic anal ysis begins. This wastes memory space and computing time. (Kita, 1992; Sano-Fukumoto, 1992) . (Aho et al., 1986) . The already existing, effi cient LR parsing algorithms can be used with TANAKA -TOKUNAGA -AIZAWA minor modifications, enabling us to utilize both the morphological and syntactic constraints at the same time. (Tomita, 1986) , but the input is not a sequence of preterminals, but a sequence of characters.", | |
| "cite_spans": [ | |
| { | |
| "start": 149, | |
| "end": 161, | |
| "text": "(Kita, 1992;", | |
| "ref_id": "BIBREF2" | |
| }, | |
| { | |
| "start": 162, | |
| "end": 182, | |
| "text": "Sano-Fukumoto, 1992)", | |
| "ref_id": "BIBREF6" | |
| }, | |
| { | |
| "start": 185, | |
| "end": 203, | |
| "text": "(Aho et al., 1986)", | |
| "ref_id": "BIBREF0" | |
| }, | |
| { | |
| "start": 414, | |
| "end": 428, | |
| "text": "(Tomita, 1986)", | |
| "ref_id": "BIBREF8" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Representing the morphological and syntacti cal constraints separately as in the first two ap proaches, Cascade and Interleave, makes main taining and extending the constraints easier. This is an advantage of these approaches . Many natu ral language processing systems have used these two approaches. For example, Mine proposed a method to represent the morphological con straints in regular grammar and the syntactic constraints in CFG, and interleave the morpho logical and syntactic analysis", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "A simple Japanese sentence consists of a sequence of postpositional phrases (PPs) followed by a predicate. The PP consists of a noun phrase (NP) followed by a postposition which indicates the case role of the NP. The predicate consists of a verb or an adjective, optionally followed by a sequence of auxiliary verbs (Morioka, 1987) . Using only the dictionary, we can obtain the 3", | |
| "cite_spans": [ | |
| { | |
| "start": 316, | |
| "end": 331, | |
| "text": "(Morioka, 1987)", | |
| "ref_id": "BIBREF5" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Morphological analysis of Japanese", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": ".", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "We illustrate the Japanese morphological analysis with an example sentence \"KaORuNi AIMaSu (meet Kaoru).\" 1 We use a simple Japanese dictionary shown in figure 2, and a con nection matrix shown in figure 3 which gives us the connectabilities between adjacent morpholog ical categories (meat). For example in figure 3, the symbol \"o\" at the intersection of row 2 (p1) and column 3 (vs4k) indicates that the morpho logical category vs4k can immediately follow the morphological category p 1.", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "Generating A Modified LR Table the sentence \"KaORuNiAIMaSu.\"", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 25, | |
| "end": 36, | |
| "text": "Table the", | |
| "ref_id": null | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "following twelve candidates of segmentation for", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "KaO Ru Ni A I MaSu (1) n1 ve4r3 p1 vs4k ve4k2i ax1", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "following twelve candidates of segmentation for", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "(2) n1 ve4r3 p1 vs4k ve4w2 ax 1 (3) n1 ve4r3 p1 vs4w ve4k2i ax 1 (4) n1 ve4r3 p1 vs4w ve4w2 ax1 (5) vs4r ve4r3 p1 vs4k ve4k2i ax 1 (6) vs4r ve4r3 p1 vs4k ve4w2 ax 1 (7) vs4r ve4r3 p1 vs4w ve4k2i ax1 (8) vs4r ve4r3 p1 vs4w ve4w2 ax 1", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "following twelve candidates of segmentation for", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "KaORu Ni A I MaSu (9) n1 p1 vs4k ve4k2i ax 1 (10) n1 p1 vs4k ve4w2 ax 1 (11) n1 p1 vs4w ve4k2i ax 1 (12) n1 p1 vs4w ve4w2 ax1", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "following twelve candidates of segmentation for", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "By also referring to the connection matrix, we can filter out illegal segmentations. From the examples above, we find (1)-(4) violate the connectability between \"KaO (n1)\" and \"Ru (ve4r3)\" , and that (5)-(8) violate the con nectability between \"Ru (ve4r3)\" and \"Ni (p1).\" Also (9) and (11) violate the connectability be tween \"I (ve4k2i)\" and \"MaSu (ax1)\", and (11) violates the connectability between \"A (vs4w)\" and \"I (ve4k2i).\" Thus by process of elimination we obtain the morphologically correct candidate, (12). However, a long input sentence generally gives many more ambiguities which need to be resolved\u2022 in later stages using syntactic, semantic and contextual information.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "following twelve candidates of segmentation for", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "Connection matrices and CFG rules have been used for morphological analysis and syntactic analysis respectively by most Japanese process ing systems. Because CFG rules were mainly used for syntactic analysis and connection matrices for morphological analysis, they have been developed independently of each other. In this section, we propose a method to inte grate morphological and syntactic constraints in the framework of LR parsing algorithm, and thus capturing the advantages of Cascade/Interleave and Single Framework described in section 1.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "following twelve candidates of segmentation for", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "In order to combine connection matrices and CFG rules, the first step we have to take is to extend the CFG rules by relating the syntactic categories in the CFG rules with the morpholog ical categories in a connection matrix. This is realized by adding CFG rules called morphologi cal rules each of which is a unit production rule with a syntactic category in the LHS and a mor phological category in the RHS.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "following twelve candidates of segmentation for", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "From the dictionary shown in figure 2, we can extract a set of new CFG rules as shown in fig ure 6 , which are simply added to the CFG rules in figure 4 to get an extended set of CFG rules with morphological constraints. ", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 89, | |
| "end": 99, | |
| "text": "fig ure 6", | |
| "ref_id": null | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "following twelve candidates of segmentation for", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "s ---+ v ax (1) v ---+ vs ve (3) s ---+ pp s (2) pp ---+ n p (4)", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "following twelve candidates of segmentation for", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "V V s V e V V e t V V V", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "ACTION GOTO", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "V a a t n p 4 4 4 k 2 r V 2 X X e 1 1 k r V 2 i 3 2 t 1 2 $ 8 v", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "ACTION GOTO", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "} } where each function is defined as follows: Rule : action -rule; returns a rule used by the reduce action.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "For each reduce action A with a morphological rule in each entry of LR table { if (Not Connect(RHS(Rule(A)), LA(A)) { delete A from the entry;", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "returns a look ahead symbols of the action.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "LA : action -symbol;", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "returns true or false with respect to the connectability of the two symbols.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Connect : symbol x symbol -{T, F};", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "returns a right hand side symbol of the rule.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "RHS : rule -symbol;", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "given an LR table and a connection matrix, this procedure can be performed automatically with out human intervention.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "The second step is to introduce the constraints on connectability into the LR table by deleting il legal reduce actions. This is carried out by mod ifying the LR table with the procedure shown in figure 7. Deleting reduce actions by applying the above procedure prohibits the application of morpho logical rules which violates the connectability be tween two adj acent words, namely the current scanned word and its lookahead word. Note that", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "It is possible to incorporate this procedure into the LR table generation process, however, it is better to keep them separate. Since this proce dure is applicable to any type of LR table, sepa rating this process from LR table generation en ables us to use the already existing LR table gen eration program.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "The second step is to introduce the constraints on connectability into the LR table by deleting il legal reduce actions. This is carried out by mod ifying the LR table with the procedure shown in figure 7. Deleting reduce actions by applying the above procedure prohibits the application of morpho logical rules which violates the connectability be tween two adj acent words, namely the current scanned word and its lookahead word. Note that", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "For example, in figure 5, the reduce action re7 in row 7 and column ve4r3 is deleted, since the connection between vs4k, the RHS of rule 7, and ve4r3, the lookahead preterminal, is prohib ited as shown in the connection matrix in figure 3 . Similarly, reduce action re7 in row 7 and column ve4w2 will be deleted and so forth. These dele tions are marked with asterisks ( *) in figure 5. Generally speaking, the size of the LR table is on the exponential order of the number of rules in the grammar. Introducing the morphological rules into the syntactic rules can cause an increase in the number of states in LR table, thereby ex ponentially increasing the size of the overall LR table in the worst case. In our method, the in crease of the number of states is equal to that of the morphological rules introduced. Suppose we add a morphological rule X -. x to the gram mar. Only the items in the form of [A -. a:", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 230, | |
| "end": 238, | |
| "text": "figure 3", | |
| "ref_id": null | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "The second step is to introduce the constraints on connectability into the LR table by deleting il legal reduce actions. This is carried out by mod ifying the LR table with the procedure shown in figure 7. Deleting reduce actions by applying the above procedure prohibits the application of morpho logical rules which violates the connectability be tween two adj acent words, namely the current scanned word and its lookahead word. Note that", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "\u2022 X ,B]", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "The second step is to introduce the constraints on connectability into the LR table by deleting il legal reduce actions. This is carried out by mod ifying the LR table with the procedure shown in figure 7. Deleting reduce actions by applying the above procedure prohibits the application of morpho logical rules which violates the connectability be tween two adj acent words, namely the current scanned word and its lookahead word. Note that", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "can produce a single new item [X -. \u2022x] from which only a single new state {[X -. x\u2022]} can be created. Thus the increase of the number of the states is equal to that of the morphological rules introduced, and the size of the LR table will not \"'-' grow exponentially.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "The second step is to introduce the constraints on connectability into the LR table by deleting il legal reduce actions. This is carried out by mod ifying the LR table with the procedure shown in figure 7. Deleting reduce actions by applying the above procedure prohibits the application of morpho logical rules which violates the connectability be tween two adj acent words, namely the current scanned word and its lookahead word. Note that", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "The LR parsing algorithm with the modified LR table is principally the same as To mita's gener alized LR parsing algorithm. The only difference is that To mita's algorithm assumes a sequence of preterminals as an input, while our algorithm as sumes a sequence of Kana characters 2 \u2022 Thus the dictionary reference process needs to be slightly modified. Figure 9 illustrates the outline of our parsing algorithm.", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 352, | |
| "end": 360, | |
| "text": "Figure 9", | |
| "ref_id": "FIGREF2" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "The second step is to introduce the constraints on connectability into the LR table by deleting il legal reduce actions. This is carried out by mod ifying the LR table with the procedure shown in figure 7. Deleting reduce actions by applying the above procedure prohibits the application of morpho logical rules which violates the connectability be tween two adj acent words, namely the current scanned word and its lookahead word. Note that", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "In figure 9 the stage number ( CS) indicates how many Kana characters have been processed. The procedure begins at stage O and ends at stage N, the length of an input sentence. In stage 0, the stack is initialized and only the node with state 0 exists (step (1)). In the outer-most loop (2)-(14), each stack top in the current stage is selected and processed. In step ( 4), the dictionary is consulted and look-ahead preterminals are obtained. An important point here is that look-ahead pretermi nals may have different Kana character lengths. A new node is introduced by a shift action at step (8) and is placed into a stage which is ahead of the current stage by the length of the look-ahead word.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "The second step is to introduce the constraints on connectability into the LR table by deleting il legal reduce actions. This is carried out by mod ifying the LR table with the procedure shown in figure 7. Deleting reduce actions by applying the above procedure prohibits the application of morpho logical rules which violates the connectability be tween two adj acent words, namely the current scanned word and its lookahead word. Note that", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "The following example well illustrates the algorithm in figure 9.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "The second step is to introduce the constraints on connectability into the LR table by deleting il legal reduce actions. This is carried out by mod ifying the LR table with the procedure shown in figure 7. Deleting reduce actions by applying the above procedure prohibits the application of morpho logical rules which violates the connectability be tween two adj acent words, namely the current scanned word and its lookahead word. Note that", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "The input sentence is \"KaORuNiAIMaSu$ (meet Kaoru).\" and we as sign position numbers between adjacent Kana characters. 0 1 2 3 4 56 7 8 9 In the following trace, the numbers in circles denote state numbers, and the numbers in squares denote the subtree number shown below the di agrams. The symbols enclosed by curly brackets denote a look ahead preterminal followed by the next applicable action, separated by a slash (/). The stage numbers are shown below the stacks. Current stage: 4 Dictionary Reference: vs4k(\"A\" ) ' at 4-5 vs4w(\"A\" ) \ufffd 4-5 Dictionary reference provides two look ahead preterminals for the next word.", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 119, | |
| "end": 137, | |
| "text": "0 1 2 3 4 56 7 8 9", | |
| "ref_id": "FIGREF2" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "The second step is to introduce the constraints on connectability into the LR table by deleting il legal reduce actions. This is carried out by mod ifying the LR table with the procedure shown in figure 7. Deleting reduce actions by applying the above procedure prohibits the application of morpho logical rules which violates the connectability be tween two adj acent words, namely the current scanned word and its lookahead word. Note that", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "We have proposed a method representing the morphological constraints in connection matrices and the syntactic constraints in CFGs, then com piling both constraints into an LR table. The compiled LR table enables us to make use of the already existing, efficient generalized LR pars ing algorithms through which integration of both \u2022 morphological and syntactic analysis is obtained .", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "Advantages of our approach can be summa rized as follows:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "\u2022 Morphological and syntactic constraints are represented separately, and it makes easier to maintain and extend them.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "\u2022 The morphological and syntactic con straints are compiled into a uniform rep resentation, an LR table. We can use the already existing efficient algorithms for gen eralized LR parsing for the analysis .", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "TANAKA -TOKUNAGA -AIZAWA", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "\u2022 Both the morphological and syntactic con straints can be used at the same time during the analysis.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "We have implemented our method using the EDR dictionary with 300,000 words (EDR, 1993) ", | |
| "cite_spans": [ | |
| { | |
| "start": 75, | |
| "end": 86, | |
| "text": "(EDR, 1993)", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "Each capitalized one or two character(s) corresponds to a single Japanese character (Kana character).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "We assume the input sentences consist of only Kana characters for brevity. Other types of characters, such as Kanji, can be also handled.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "If two stack tops are merged and then different shift actions (sh16 and sh18) are carried out, we might have invalid combinations of structure such as (14, 17) and (15, 16).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| } | |
| ], | |
| "back_matter": [ | |
| { | |
| "text": "After the two reduce actions (re6), we get two nodes with the same state 20, but they are not merged as the look .ahead preterminals are differ ent each other. See stage 5 for the reason.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "annex", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "The process in stage 4 continues as follows. Dictionary Reference: ve4k2i(\"I\" ) at 5-6 ve4w2( \"I\" ) at 5-6We have two look ahead preterminals and two stack tops. The reduce actions (re7 and re9) are performed.{ ve4k2i/re7 } = 3 12 7 ve4v2/errNote that we can not merge the stack tops with the same state 4 since the look ahead pretermi nals are different (ve4k2i/ve4w2 Current stage: 6 Dictionary reference: ax 1(\"MaSu\") at 6-8The process in stage 6 proceeds as follows. ", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "rill--@ {vs4k/re4}", | |
| "sec_num": null | |
| } | |
| ], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "Compilers Principles, Techniques, and To ols", | |
| "authors": [ | |
| { | |
| "first": "A", | |
| "middle": [ | |
| "V" | |
| ], | |
| "last": "Aho", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Sethi", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [ | |
| "D" | |
| ], | |
| "last": "Ullman", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1986, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Aho, A.V. -Sethi,R. -Ullman, J.D. (1986) Compilers Principles, Techniques, and To ols. Massachusetts: Addison-We sley.", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "Japan Electronic Dictionary Research Institute (1993 ) EDR Dictionary Manual", | |
| "authors": [], | |
| "year": null, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Japan Electronic Dictionary Research Institute (1993 ) EDR Dictionary Manual.", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "A Study on Language Modeling fo r Speech Recognition", | |
| "authors": [ | |
| { | |
| "first": "K", | |
| "middle": [], | |
| "last": "Kita", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1992, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Kita, K. (1992) A Study on Language Modeling fo r Speech Recognition. PhD thesis, Was eda University.", | |
| "links": null | |
| }, | |
| "BIBREF3": { | |
| "ref_id": "b3", | |
| "title": "A parallel syntactic analysis of context free grammars", | |
| "authors": [ | |
| { | |
| "first": "T", | |
| "middle": [], | |
| "last": "Mine", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Taniguchi", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Amamiya", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1990, | |
| "venue": "the 40th Annual Convention IPS Japan", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "452--453", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Mine,T. -Taniguchi, R. -Amamiya, M. (1990) \"A parallel syntactic analysis of context free grammars.\" pp. 452-453. the 40th Annual Convention IPS Japan.", | |
| "links": null | |
| }, | |
| "BIBREF4": { | |
| "ref_id": "b4", | |
| "title": "An Automatic Segmenta tion Method for Compound Wo rds using De pendency Analysis", | |
| "authors": [ | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Miyazaki", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1984, | |
| "venue": "Transactions of Information Processing Society of Japan", | |
| "volume": "25", | |
| "issue": "6", | |
| "pages": "970--979", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Miyazaki, M. (1984) \"An Automatic Segmenta tion Method for Compound Wo rds using De pendency Analysis.\" Transactions of Infor- mation Processing Society of Japan. Vol. 25, No. 6, pp. 970-979", | |
| "links": null | |
| }, | |
| "BIBREF5": { | |
| "ref_id": "b5", | |
| "title": "Fo rmation of a Vo cabulary", | |
| "authors": [ | |
| { | |
| "first": "K", | |
| "middle": [], | |
| "last": "Morioka", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1987, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Morioka, K. (1987) Fo rmation of a Vo cabulary. Meiji-Shoin.", | |
| "links": null | |
| }, | |
| "BIBREF6": { | |
| "ref_id": "b6", | |
| "title": "On a Gram mar Formalism, Knowledge Bases and Tools for Natural Language Processing in Logic Pro gramming", | |
| "authors": [ | |
| { | |
| "first": "H", | |
| "middle": [], | |
| "last": "Sano", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "F", | |
| "middle": [], | |
| "last": "Fukumoto", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1992, | |
| "venue": "Proceedings of FGCS92", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Sano, H. -Fukumoto, F. (1992) \"On a Gram mar Formalism, Knowledge Bases and Tools for Natural Language Processing in Logic Pro gramming.\" in Proceedings of FGCS92.", | |
| "links": null | |
| }, | |
| "BIBREF7": { | |
| "ref_id": "b7", | |
| "title": "Logic Programming '88 (Lecture Notes in Artificial Intelligence)", | |
| "authors": [ | |
| { | |
| "first": "R", | |
| "middle": [], | |
| "last": "Sugimura", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "K", | |
| "middle": [], | |
| "last": "Akasaka", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Y. -Mat", | |
| "middle": [], | |
| "last": "Kubo", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Y", | |
| "middle": [], | |
| "last": "Sumoto", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1989, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "188--216", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Sugimura,R. -Akasaka, K. -Kubo, Y. -Mat sumoto, Y. (1989) \"Logic Based Lexical Ana lyzer LAX.\" Logic Programming '88 (Lecture Notes in Artificial Intelligence). pp. 188-216. Springer-Verlag.", | |
| "links": null | |
| }, | |
| "BIBREF8": { | |
| "ref_id": "b8", | |
| "title": "Effi cient Parsing fo r Natu ral Language: A Fast Algorithm fo r Practical Systems", | |
| "authors": [ | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Tomita", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1986, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Tomita, M. (1986) Effi cient Parsing fo r Natu ral Language: A Fast Algorithm fo r Practical Systems. Boston: Kluwer Academic Publish ers.", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": { | |
| "FIGREF0": { | |
| "text": "Fig. 2 A simple Japanese dictionary Fig. 3 An example of connection matrix", | |
| "uris": null, | |
| "type_str": "figure", | |
| "num": null | |
| }, | |
| "FIGREF1": { | |
| "text": "Fig. 6 A morphological rules derived from the dictionary in Fig. 2 We can generate an LR table as shown in figure 5 from the extended CFG rules (1) through (16) from figure 4 and 6. Note that the extended CFG rules do not include any in formation about connectability represented in the connection matrix in figure 3. For exam ple, rules (3), (8) and (13) allow the struc ture \"v ( vs ( vs4r) , ve ( ve4w2) )\" which violates the connectability between vs4r and ve4w2 as shown in figure 3.", | |
| "uris": null, | |
| "type_str": "figure", | |
| "num": null | |
| }, | |
| "FIGREF2": { | |
| "text": "Outline of our parsing algorithmThe overview of generating a modified LR table 4is shown in figure 8.", | |
| "uris": null, | |
| "type_str": "figure", | |
| "num": null | |
| }, | |
| "FIGREF3": { | |
| "text": "look ahead preterminals, n1, vs4r, and n1 by consulting the dictionary in figure 2. A shift actions is applied for each of them according to the LR table in figure 5.@ {n1/sh6 , vs4r/sh8 , n1/sh6} After the shift actions, three new nodes are cre ated at stage 2 or stage 3 depending on the length of look ahead words. At the same time subtrees 1-3 are constructed. The current stage is updated from O to 2, since there is no node in stage 1.The look ahead preterminals are unknown at this mo men t. the first two stack tops are con cerned at this stage. No action is taken of the first stack, because the LR table has no action in the entry for state 6 and a look ahead preterminal ve4r3. As the result, the first stack is rejected.The reduce action (re8) is taken for the second stack.", | |
| "uris": null, | |
| "type_str": "figure", | |
| "num": null | |
| }, | |
| "TABREF2": { | |
| "html": null, | |
| "content": "<table><tr><td>\ufffd Pr 1 1</td><td>\ufffd v0ie i x</td></tr><tr><td>n1 I KaORu Ni p1 I</td><td>vs4w ve4w2 ax1 I I I A I MaSu</td></tr></table>", | |
| "type_str": "table", | |
| "text": "", | |
| "num": null | |
| } | |
| } | |
| } | |
| } |