A Multi-Strategy Approach to Parsing of Grammatical Relations in Child Language Transcripts Kenji Sagae Language Technologies Institute Carnegie Mellon University Thesis Committee: Alon Lavie, co-chair Brian MacWhinney, co-chair Lori Levin Jaime Carbonell John Carroll, University of Sussex Natural Language Parsing: Sentence → Syntactic Structure • One of the core problems in NLP Input: The boy ate the cheese sandwich Output: (ROOT (predicate eat) (surface ate) (tense past) (category V) Grammatical Relations (GRs) ((1 (S 2(NPThe (Det DET)N) (agreement 3s) (SUBJThe) (category • Subject, object, adjunct, etc. (surface boy) (2 3 boy (N boy)) SUBJ) (DET (surface the) (3 0(VPate (V ate) ROOT) (category Det))) (category N) (definite +) (4 6 the (NP(OBJ (Det the) DET) (DET (surface the) (5 6 cheese(N cheese) MOD) (category Det)) sandwich) (6 3 sandwich (N(predicate sandwich)))) OBJ)) (surface sandwich) (MOD (category N) (surface cheese) (predicate cheese)))) 2 Using Natural Language Processing in Child Language Research • CHILDES Database (MacWhinney, 2000) – 200 megabytes of child-parent dialog transcripts – Part-of-speech and morphology analysis • Tools available • Not enough for many research questions – No syntactic analysis • Can we use NLP to analyze CHILDES transcripts? – Parsing – Many decisions: representation, approach, etc. 3 Parsing CHILDES: Specific and General Motivation • Specific task: automatic analysis of syntax in CHILDES corpora – Theoretical importance (study of child language development) – practical importance (measurement of syntactic competence) • In general: Develop techniques for syntactic analysis, advance parsing technologies – Can we develop new techniques that perform better than current approaches? • Rule-based • Data-driven 4 Research Objectives • Identify a suitable syntactic representation for CHILDES transcripts – Must address the needs of child language research • Develop a high accuracy approach for syntactic analysis of spoken language transcripts – parents and children at different stages of language acquisition • The plan: a multi-strategy approach – ML: ensemble methods – Parsing: several approaches possible, but combination is an underdeveloped area 5 Research Objectives • Develop methods for combining analyses from different parsers and obtain improved accuracy – Combining rule-based and data-driven approaches • Evaluate the accuracy of developed systems • Validate the utility of the resulting systems to the child language community – Task-based evaluation: Automatic measurement of grammatical complexity in child language 6 Overview of the Multi-Strategy Approach for Syntactic Analysis Parser A Parser B Transcripts Parser Combination Parser C Parser D SYNTACTIC STRUCTURES Parser E 7 Thesis Statement • The development of a novel multi-strategy approach for syntactic parsing allows for identification of Grammatical Relations in transcripts of parent-child dialogs at a higher level of accuracy than previously possible • Through the combination of different NLP techniques (rule-based or data-driven), the multi-strategy approach can outperform each strategy in isolation, and produce significantly improved accuracy • The resulting syntactic analysis are at a level of accuracy that makes them useful to child language research 8 Outline • The CHILDES GR scheme • GR Parsing of CHILDES transcripts • Combining different strategies • Automated measurement of syntactic development in child language • Related work • Conclusion 9 CHILDES GR Scheme (Sagae, MacWhinney and Lavie, 2004) • Grammatical Relations (GRs) – Subject, object, adjunct, etc. – Labeled dependencies • Addresses needs of child language researchers – Informative and intuitive, basis for DSS and IPSyn Dependency Label Dependent Head 10 CHILDES GR Scheme Includes Important GRs for Child Language Study 11 Outline • The CHILDES GR scheme • GR Parsing of CHILDES transcripts • Evaluation • Data • Combining different strategies • Rule-based GR parsing • Data-driven GR parsing • Automated measurement of syntactic development in child language • Related word • Conclusion 12 The Task: Sentence → GRs • Input: We eat the cheese sandwich • Output: 13 Evaluation of GR Parsing • Dependency accuracy • Precision/Recall of GRs 14 Evaluation: Calculating Dependency Accuracy 1 2 3 4 5 1 2 3 2 0 5 5 2 We eat the cheese sandwich 4 5 SUBJ ROOT DET MOD SUBJ 15 Evaluation: Calculating Dependency Accuracy GOLD 1 PARSED 2 3 1 2 3 4 4 5 2 0 4 2 2 We eat the cheese 5 sandwich Accuracy =1number of correct dependencies 2 We SUBJ of dependencies 2 total 0 number eat ROOT 3 5 the DET = 2 /45 =50.40cheese MOD 5 2 sandwich SUBJ 40% SUBJ ROOT DET OBJ PRED 16 Evaluation: Precision and Recall of GRs • Precision and recall are calculated separately for each GR type • Calculated on aggregate counts over entire test corpus • Example: SUBJ Precision = # SUBJ matches between PARSED and GOLD Total number of SUBJs in PARSED Recall = # SUBJ matches between PARSED and GOLD Total # of SUBJs in GOLD F-score = 2 ( Precision × Recall ) Precision + Recall 17 Evaluation: Precision and Recall of GRs GOLD 1 2 3 4 5 2 0 5 5 2 We eat the cheese sandwich PARSED SUBJ ROOT DET MOD OBJ 1 2 3 4 5 2 0 4 2 2 We eat the cheese sandwich SUBJ ROOT DET OBJ SUBJ Precision SUBJ matches between PARSED and and GOLD GOLD F-score Recall === ##2SUBJ ( Precision matches × Recall between ) PARSED Total number of SUBJs in PARSED Precision + Recall Total # of SUBJs in GOLD == 1 2(50×100) 1 // 2 1 = = 50% 100% / (50+100) = 66.67 = 18 Outline • The CHILDES GR scheme • GR Parsing of CHILDES transcripts • • • • Evaluation Data Rule-based GR parsing Data-driven GR parsing • Combining different strategies • Automated measurement of syntactic development in child language 19 CHILDES Data: the Eve Corpus (Brown, 1973) • A corpus from CHILDES – Manually annotated with GRs • Training: ~ 5,000 words (adult) • Development: ~ 1,000 words – 600 adult, 400 child • Test: ~ 2,000 words – 1,200 adult, 800 child 20 Not All Child Utterances Have GRs • Utterances in training and test sets are well-formed I need tapioca in the bowl. That’s a hat. In a minute. • What about * Warm puppy happiness a blanket. * There briefcase. ? I drinking milk. ? I want Fraser hat. • Separate Eve-child test set (700 words) 21 The WSJ Corpus (Penn Treebank) • 1 million words • Widely used – Sections 02-21: training – Section 22: development – Section 23: evaluation • Large corpus with syntactic annotation – Out-of-domain • Constituent structures – Convert to unlabeled dependencies using headpercolation table 22 Outline • The CHILDES GR scheme • GR Parsing of CHILDES transcripts • • • • Evaluation Data Rule-based GR parsing Data-driven GR parsing • Combining different strategies • Automated measurement of syntactic development in child language 23 Rule-Based Parsing • The parser’s knowledge is encoded in manually written rules – Grammar, lexicon, etc. • Only analyses that fit the rules are possible • Accurate in specific domains, difficult to achieve wide coverage in open domain – Coverage, ambiguity, domain knowledge 24 Rule-Based Parsing of CHILDES data (Sagae, Lavie & MacWhinney, 2001, 2004) LCFlex (Rosé and Lavie, 2001) Rules: CFG backbone augmented with unification constraints Manually written, 153 rules Robustness Limited insertions: [Do] [you] want to go outside? Limited skipping: No um maybe later. PCFG disambiguation model Trained on 2,000 words 25 High Precision from a Small Grammar • Eve test corpus – 2,000 words • • • • 31% of the words can be parsed Accuracy (over all 2,000 words): 29% Precision: 94% High Precision, Low Recall • Improve recall using parser’s robustness – Insertions, skipping – Multi-pass approach 26 Robustness and Multi-Pass Parsing • No insertions, no skipping 31% parsed, 29% recall, 94% precision • Insertion of NP and/or auxiliary 38% parsed, 35% recall, 92% precision • Skipping of 1 word 52% parsed, 47% recall, 90% precision • Skipping of 1 word, insertion of NP, aux 63% parsed, 55% recall, 88% precision 27 Use Robustness to Improve Recall 100 90 80 70 60 precision 50 recall 40 f-score 30 20 10 0 none insert NP/aux skip 1 word insert/skip 28 Outline • The CHILDES GR scheme • GR Parsing of CHILDES transcripts • • • • Evaluation Data Rule-based GR parsing Data-driven GR parsing • Combining different strategies • Automated measurement of syntactic development in child language 29 Data-driven Parsing • Parser learns from a corpus of annotated examples • Data-driven parsers are robust • Two approaches – Existing statistical parser – Classifier-based parsing 30 Accurate GR Parsing with Existing Resources (Mostly) • Large training corpus: Penn Treebank (Marcus et al., 1993) – Head-table converts constituents into dependencies • Use an existing parser (trained on the Penn Treebank) – Charniak (2000) • Convert output to unlabeled dependencies • Use a classifier for dependency labeling 31 Unlabeled Dependency Identification We eat the cheese sandwich eat eat sandwich 32 Domain Issues • Parser training data is in a very different domain – WSJ vs Parent-child dialogs • Domain specific training data would likely be better • Performance is acceptable – Shorter, simpler sentences – Unlabeled dependency accuracy • WSJ test data: 92% • Eve test data: 90% 33 Dependency Labeling • Training data is required – Eve training set (5,000 words) • Labeling dependencies is easier than finding unlabeled dependencies • Use a classifier – TiMBL (Daelemans et al., 2004) – Extract features from unlabeled dependency structure – GR labels are target classes 34 Dependency Labeling 35 Features Used for GR Labeling • Head and dependent words – Also their POS tags • Whether the dependent comes before or after the head • How far the dependent is from the head • The label of the lowest node in the constituent tree that includes both the head and dependent 36 Features Used for GR Labeling Consider the words “we” and “eat” Features: we, pro, eat, v, before, 1, S Class: SUBJ 37 Good GR Labeling Results with Small Training Set • Eve training set – 5,000 words for training • Eve test set – 2,000 words for testing • Accuracy of dependency labeling (on perfect dependencies): 91.4% • Overall accuracy (Charniak parser + dependency labeling): 86.9% 38 Some GRs Are Easier Than Others • Overall accuracy: 86.9% • Easily identifiable GRs – DET, POBJ, INF, NEG: Precision and recall above 98% • Difficult GRs – COMP, XCOMP: below 65% • I think that Mary saw a movie (COMP) • She tried to see a movie (XCOMP) 39 Precision and Recall of Specific GRs GR SUBJ Precision 0.94 Recall 0.93 F-score 0.93 OBJ COORD JCT 0.83 0.68 0.91 0.91 0.85 0.82 0.87 0.75 0.86 MOD PRED ROOT 0.79 0.80 0.91 0.92 0.83 0.92 0.85 0.81 0.91 COMP XCOMP 0.60 0.58 0.50 0.64 0.54 0.61 40 Parsing with Domain-Specific Data • Good results with a system based on the Charniak parser • Why domain-specific data? – No Penn Treebank – Handle dependencies natively – Multi-strategy approach 41 Classifier-Based Parsing (Sagae & Lavie, 2005) • Deterministic parsing – Single path, no backtracking – Greedy – Linear run-time • Simple shift-reduce algorithm – Single pass over the input string • Variety: Left-to-right, right-to-left (order matters) • Classifier makes parser decisions – Classifier not tied to parsing algorithm • Variety: Different types of classifiers can be used 42 A Simple, Fast and Accurate Approach • Classifier-based parsing with constituents – Trained and evaluated on WSJ data: 87.5% – Very fast, competitive accuracy • Simple adaptation to labeled dependency parsing – Similar to Malt parser (Nivre, 2004) – Handles CHILDES GRs directly 43 GR Analysis with Classifier-Based Parsing • Stack S – Items may be POS-tagged words or dependency trees – Initialization: empty • Queue W – Items are POS-tagged words – Initialization: Insert each word of the input sentence in order (first word is in front) 44 Shift and Reduce Actions • Shift – Remove (shift) the word in front of queue W – Insert shifted item on top of stack S • Reduce – Pop two topmost item from stack S – Push new item onto stack S • New item forms new dependency • Choose LEFT or RIGHT • Choose Dependency Label 45 Parser Decisions • Shift vs. Reduce • If Reduce – RIGHT or LEFT – Dependency label • We use a classifier to make these decisions 46 Classes and Features • Classes – – – – – – SHIFT LEFT-SUBJ LEFT-JCT RIGHT-OBJ RIGHT-JCT … • Features: derived from parser configuration – Crucially: two topmost items in S, first item in W – Additionally: other features that describe the current configuration (look-ahead, etc) 47 Parsing CHILDES with a Classifier-Based Parser • Parser uses SVM • Trained on Eve training set (5,000 words) • Tested on Eve test set (2,000 words) • Labeled dependency accuracy: 87.3% – Uses only domain-specific data – Same level of accuracy as GR system based on Charniak parser 48 Precision and Recall of Specific GRs GR SUBJ Precision 0.97 Recall 0.98 F-score 0.98 OBJ COORD JCT 0.89 0.71 0.78 0.94 0.76 0.88 0.92 0.74 0.83 MOD PRED ROOT 0.94 0.80 0.95 0.87 0.83 0.94 0.91 0.82 0.94 COMP XCOMP 0.70 0.93 0.78 0.82 0.74 0.87 49 Precision and Recall of Specific GRs GR SUBJ Precision 0.97 Recall 0.98 F-score 0.98 0.93 OBJ COORD JCT 0.89 0.71 0.78 0.94 0.76 0.88 0.92 0.87 0.74 0.75 0.83 0.86 MOD PRED ROOT 0.94 0.80 0.95 0.87 0.83 0.94 0.91 0.85 0.82 0.81 0.94 0.91 COMP XCOMP 0.70 0.93 0.78 0.82 0.74 0.54 0.87 0.61 50 Outline • The CHILDES GR scheme • GR Parsing of CHILDES transcripts • Combining different strategies • Weighted voting • Automated measurement • Combination as parsing of syntactic • Handling young childlanguage utterances development in child • Related Work • Conclusion 51 Combine Different Parsers to Get More Accurate Results • Rule-based • Statistical parsing + dependency labeling • Classifier-based parsing – Obtain even more variety • SVM vs MBL • Left-to-right vs right-to-left 52 Simple (Unweighted) Voting • Each parser votes for each dependency • Word-by-word • Every vote has the same weight 53 Simple (Unweighted) Voting He eats cake Parser A 1 2 He SUBJ 2 0 eats CMOD 3 1 cake OBJ Parser B Parser C 1 2 He SUBJ 1 3 He SUBJ 2 0 eats ROOT 2 0 eats ROOT 3 1 cake OBJ 3 2 cake OBJ GOLD 1 2 He SUBJ 2 0 eats ROOT 3 2 cake OBJ 54 Simple (Unweighted) Voting He eats cake Parser A 1 2 He SUBJ 2 0 eats CMOD 3 1 cake OBJ GOLD 1 2 He SUBJ 2 0 eats ROOT 3 2 cake OBJ Parser B Parser C 1 2 He SUBJ 1 3 He SUBJ 2 0 eats ROOT 2 0 eats ROOT 3 1 cake OBJ 3 2 cake OBJ VOTED 1 2 He SUBJ 2 0 eats ROOT 3 1 cake SUBJ 55 Simple (Unweighted) Voting He eats cake Parser A 1 2 He SUBJ 2 0 eats CMOD 3 1 cake OBJ GOLD 1 2 He SUBJ 2 0 eats ROOT 3 2 cake OBJ Parser B Parser C 1 2 He SUBJ 1 3 He SUBJ 2 0 eats ROOT 2 0 eats ROOT 3 1 cake OBJ 3 2 cake OBJ VOTED 1 2 He SUBJ 2 0 eats ROOT 3 1 cake OBJ 56 Weighted Voting • Each parser has a weight – Reflects confidence in parser’s GR identification • Instead of adding number of votes, add the weight of votes • Takes into account that some parsers are better than others 57 Weighted Voting He eats cake Parser A (0.4) 1 2 He SUBJ 2 0 eats CMOD 3 1 cake OBJ GOLD 1 2 He SUBJ 2 0 eats ROOT 3 2 cake OBJ Parser B (0.3) Parser C (0.8) 1 2 He SUBJ 1 3 He SUBJ 2 0 eats ROOT 2 0 eats ROOT 3 1 cake OBJ 3 2 cake OBJ VOTED 1 3 He SUBJ 2 0 eats ROOT 3 2 cake OBJ 58 Label-Weighted Voting • Not just one weight per parser, but one weight for each GR for each parser • Takes into account specific strengths of each parser 59 Label-Weighted Voting He eats cake Parser A 1 2 He Parser B SUBJ (0.7) 1 2 He Parser C SUBJ (0.8) 1 3 He SUBJ (0.6) 2 0 eats CMOD (0.3) 2 0 eats ROOT (0.9) 2 0 eats ROOT(0.7) 3 1 cake OBJ (0.5) GOLD 1 2 He SUBJ 2 0 eats ROOT 3 2 cake OBJ 3 1 cake OBJ (0.3) 3 2 cake OBJ (0.9) VOTED 1 2 He SUBJ 2 0 eats ROOT 3 2 cake OBJ 60 Voting Produces Very Accurate Results • Parsers – Rule-based – Statistical based on Charniak parser – Classifier-based • Left-to-right SVM • Right-to-left SVM • Left-to-right MBL • Simple Voting: 88.0% • Weighted Voting: 89.1% • Label-weighted Voting: 92.1% 61 Precision and Recall of Specific GRs GR SUBJ Precision 0.98 Recall 0.98 F-score 0.98 OBJ COORD JCT 0.94 0.94 0.87 0.94 0.91 0.90 0.94 0.92 0.88 MOD PRED ROOT 0.97 0.86 0.97 0.91 0.89 0.96 0.94 0.87 0.96 COMP XCOMP 0.75 0.90 0.67 0.88 0.71 0.89 62 Outline • The CHILDES GR scheme • GR Parsing of CHILDES transcripts • Combining different strategies • Weighted voting • Combination as parsing • Handling young child utterances • Automated measurement of syntactic development in child language 63 Voting May Not Produce a Well-Formed Dependency Tree • Voting on a word-by-word basis • No guarantee of well-formedness • Resulting set of dependencies may form a graph with cycles, or may not even be fully connected – Technically not fully compliant with CHILDES GR annotation scheme 64 Parser Combination as Reparsing • Once several parsers have analyzed a sentence, use their output to guide the process of reparsing the sentence • Two reparsing approaches – Maximum spanning tree – CYK (dynamic programming) 65 Dependency Parsing as Search for Maximum Spanning Tree • First, build a graph – Each word in input sentence is a node – Each dependency proposed by any of the parsers is an weighted edge – If multiple parsers propose the same dependency, add weights into a single edge • Then, simply find the MST – Maximizes the votes – Structure guaranteed to be a dependency tree – May have crossing branches 66 Parser Combination with the CYK Algorithm • The CYK algorithm uses dynamic programming to find all parses for a sentence given a CFG – Probabilistic version finds most probable parse • Build a graph, as with MST • Parse the sentence using CYK – Instead of a grammar, consult the graph to determine how to fill new cells in the CYK table – Instead of probabilities, we use the weights from the graph 67 Precision and Recall of Specific GRs GR SUBJ Precision 0.98 Recall 0.98 F-score 0.98 OBJ COORD JCT 0.94 0.94 0.87 0.94 0.91 0.90 0.94 0.92 0.88 MOD PRED ROOT 0.97 0.86 0.97 0.91 0.89 0.97 0.94 0.87 0.97 COMP XCOMP 0.73 0.88 0.89 0.88 0.80 0.88 68 Outline • The CHILDES GR scheme • GR Parsing of CHILDES transcripts • Combining different strategies • Weighted voting • Combination as parsing • Handling young child utterances • Automated measurement of syntactic development in child language 69 Handling Young Child Utterances with Rule-Based and Data-Driven Parsing • Eve-child test set: I need tapioca in the bowl. That’s a hat. In a minute. * Warm puppy happiness a blanket. * There briefcase. ? I drinking milk. ? I want Fraser hat. 70 Three Types of Sentences in One Corpus • No problem – High accuracy • No GRs – But data-driven systems will output GRs • Missing words, agreement errors, etc. – GRs are fine, but a challenge for data-driven systems trained on fully grammatical utterances 71 To Analyze or Not To Analyze: Ask the Rule-Based Parser • Utterances with no GRs are annotated in test corpus as such • Rule-based parser set to high precision – Same grammar as before • If sentence cannot be parsed with the rulebased system, output No GR. – 88% Precision, 89% Recall – Sentences are fairly simple 72 The Rule-Based Parser also Identifies Missing Words • If the sentence can be analyzed with the rule-based system, check if any insertions were necessary – If inserted be or possessive marker ’s, insert the appropriate lexical item in the sentence • Parse the sentence with data-driven systems, run combination 73 High Accuracy Analysis of Challenging Utterances • Eve-child test – No rule-based first pass: 62.9% accuracy • Many errors caused by GR analysis of words with no GRs – With rule-based pass: 88.0% accuracy • 700 words from Naomi corpus – No rule-based: 67.4% – Rule-based, then combo: 86.8% 74 Outline • The CHILDES GR scheme • GR Parsing of CHILDES transcripts • Combining different strategies • Automated measurement of syntactic development in child language • Related work • Conclusion 75 Index of Productive Syntax (IPSyn) (Scarborough, 1990) • A measure of child language development • Assigns a numerical score for grammatical complexity (from 0 to 112 points) • Used in hundreds of studies 76 IPSyn Measures Syntactic Development • IPSyn: Designed for investigating differences in language acquisition – Differences in groups (for example: bilingual children) – Individual differences (for example: delayed language development) – Focus on syntax • Addresses weaknesses of Mean Length of Utterance (MLU) – MLU surprisingly useful until age 3, then reaches ceiling (or becomes unreliable) • IPSyn is very time-consuming to compute 77 Computing IPSyn (manually) • Corpus of 100 transcribed utterances – Consecutive, no repetitions • Identify 56 specific language structures (IPSyn Items) – Examples: • • • • Presence of auxiliaries or modals Inverted auxiliary in a wh-question Conjoined clauses Fronted or center-embedded subordinate clauses – Count occurrences (zero, one, two or more) • Add counts 78 Automating IPSyn • Existing state of manual computation – Spreadsheets – Search each sentence for language structures – Use part-of-speech tagging to narrow down the number of sentences for certain structures • For example: Verb + Noun, Determiner + Adjective + Noun • Automatic computation is possible with accurate GR analysis – Use GRs to search for IPSyn items 79 Some IPSyn Items Require Syntactic Analysis for Reliable Recognition (and some don’t) • • • • • • • • • • • Determiner + Adjective + Noun Auxiliary verb Adverb modifying adjective or nominal Subject + Verb + Object Sentence with 3 clauses Conjoined sentences Wh-question with inverted auxiliary/modal/copula Relative clauses Propositional complements Fronted subordinate clauses Center-embedded clauses 80 Automating IPSyn with Grammatical Relation Analyses • Search for language structures using patterns that involve POS tags and GRs (labeled dependencies) • Examples – Wh-embedded clauses: search for wh-words whose head (or transitive head) is a dependent in a GR of types [XC]SUBJ, [XC]PRED, [XC]JCT, [XC]MOD, COMP or XCOMP – Relative clauses: search for a CMOD where the dependent is to the right of the head 81 Evaluation Data • Two sets of transcripts with IPSyn scoring from two different child language research groups • Set A – Scored fully manually – 20 transcripts – Ages: about 3 yrs. • Set B – Scored with CP first, then manually corrected – 25 transcripts – Ages: about 8 yrs. (Two transcripts in each set were held out for development and debugging) 82 Evaluation Metrics: Point Difference • Point difference – The absolute point difference between the scores provided by our system, and the scores computed manually – Simple, and shows how close the automatic scores are to the manual scores – Acceptable range • Smaller for older children 83 Evaluation Metrics: Point-to-Point Accuracy • Point-to-point accuracy – Reflects overall reliability over each scoring decision made in the computation of IPSyn scores – Scoring decisions: presence or absence of language structures in the transcript Point-to-Point Acc = C(Correct Decisions) C(Total Decisions) – Commonly used for assessing inter-rater reliability among human scorers (for IPSyn, about 94%). 84 Results • IPSyn scores from – Our GR-based system (GR) – Manual scoring (HUMAN) – Computerized Profiling (CP) • Long, Fey and Channell, 2004 85 GR-based IPSyn Is Quite Accurate System Avg. Point Difference Point-to-point to HUMAN Reliability (%) GR (total) 3.3 92.8 CP (total) 8.3 85.4 GR (set A) 3.7 92.5 CP (set A) 6.2 86.2 GR (set B) 2.9 93.0 CP (set B) 10.2 84.8 86 GR-Based IPSyn Close to Human Scoring • Automatic scores very reliable • Validates usefulness of – GR annotation scheme – Automatic GR analysis • Validates analysis over a large set of children of different ages 87 Outline • The CHILDES GR scheme • GR Parsing of CHILDES transcripts • Combining different strategies • Automated measurement of syntactic development in child language • Related work • Conclusion 88 Related Work • GR schemes, GR evaluation: – Carroll, Briscoe & Sanfilippo, 1998 – Lin, 1998 – Yeh, 2000 – Preiss, 2003 • Rule-based robust parsing – Heeman & Allen, 2001 – Lavie, 1996 – Rosé & Lavie, 2001 • Parsing – Carroll & Briscoe, 2002 – Briscoe & Carroll, 2002 – Buchholz, 2002 – Tomita, 1987 – Magerman, 1995 – Ratnaparkhi, 1997 – Collins, 1997 – Charniak, 2000 • Deterministic parsing – Yamada & Matsumoto, 2003 – Nivre & Scholz, 2004 • Parser Combination – Henderson & Brill, 1999 – Brill & Wu, 1998 – Yeh, 2000 – Sarkar, 2001 • Automatic measurement of grammatical complexity – Long, Fey & Channell, 89 2004 Outline • The CHILDES GR scheme • GR Parsing of CHILDES transcripts • Combining different strategies • Automated measurement of syntactic development in child language • Related work • Conclusion 90 Major Contributions • An annotation scheme based on GRs for syntactic structure in CHILDES transcripts • A linear-time classifier-based parser for constituent structures • The development of rule-based and data-driven approaches to GR analysis – Precision/recall trade-off using insertions and skipping – Data-driven GR analysis using existing resources • Charniak parser, Penn Treebank – Parser variety in classifier-based dependency parsing 91 Major Contributions (2) • The use of different voting schemes for combining dependency analyses – Surpasses state-of-the-art in WSJ dependency parsing – Vastly outperforms individual parsing approaches • A novel reparsing combination scheme – Maximum spanning trees, CYK • An accurate automated tool for measurement of syntactic development in child language – Validates annotation scheme and quality of GR analyses 92 Possible Future Directions • Classifier-based parsing – Beam search keeping linear time – Tree classification (Kudo & Matsumoto, 2004) • Parser combination – Parser variety, reparsing combination with constituent trees • Automated measurement of grammatical complexity – Take precision/recall into account – A data-driven approach to replace search rules • Other languages 93 94 95 96 97 More on Dependency Voting • On WSJ data: 93.9% unlabeled accuracy • On Eve data – No RB: 91.1% • COMP: 50% – No charn, No RB: 89.1% • COMP: 50%, COORD: 84%, ROOT: 95% – No charn: 90.5% • COMP: 67% – No RL, no MBL: 91.8% 98 Full GR Results • • • • • • • • • • • • • • • • • • • • • • • • • XJCT ( 2 / 2) : 1.00 1.00 1.00 OBJ ( 90 / 91) : 0.95 0.96 0.95 NEG ( 26 / 25) : 1.00 0.96 0.98 SUBJ ( 180 / 181) : 0.98 0.98 0.98 INF ( 19 / 19) : 1.00 1.00 1.00 POBJ ( 48 / 51) : 0.92 0.98 0.95 XCOMP ( 23 / 23) : 0.88 0.88 0.88 QUANT ( 4 / 4) : 1.00 1.00 1.00 VOC ( 2 / 2) : 1.00 1.00 1.00 TAG ( 1 / 1) : 1.00 1.00 1.00 CPZR ( 10 / 9) : 1.00 0.90 0.95 PTL ( 6 / 6) : 0.83 0.83 0.83 COORD ( 33 / 33) : 0.91 0.91 0.91 COMP ( 18 / 18) : 0.71 0.89 0.80 AUX ( 74 / 78) : 0.94 0.99 0.96 CJCT ( 6 / 5) : 1.00 0.83 0.91 PRED ( 54 / 55) : 0.87 0.89 0.88 DET ( 45 / 47) : 0.96 1.00 0.98 MOD ( 94 / 89) : 0.97 0.91 0.94 ROOT ( 239 / 238) : 0.97 0.96 0.96 PUNCT ( 286 / 286) : 1.00 1.00 1.00 COM ( 45 / 44) : 0.93 0.91 0.92 ESUBJ ( 2 / 2) : 1.00 1.00 1.00 CMOD ( 3 / 3) : 1.00 1.00 1.00 JCT ( 78 / 84) : 0.85 0.91 0.88 99 Weighted Voting 0.7 Parser A (0.4) 1 2 He SUBJ 2 0 eats CMOD 3 1 cake OBJ GOLD 1 2 He SUBJ 2 0 eats ROOT 3 2 cake OBJ He eats cake 0.8 Parser B (0.3) Parser C (0.8) 1 2 He SUBJ 1 3 He SUBJ 2 0 eats ROOT 2 0 eats ROOT 3 1 cake OBJ 3 2 cake OBJ VOTED 1 3 He SUBJ 2 0 eats ROOT 3 2 cake OBJ 100 Weighted Voting 1.5 Parser A 1 2 He He eats cake Parser B SUBJ (0.7) 1 2 He 0.6 Parser C SUBJ (0.8) 1 3 He SUBJ (0.6) 2 0 eats CMOD (0.3) 2 0 eats ROOT (0.9) 2 0 eats ROOT(0.7) 3 1 cake OBJ (0.5) GOLD 1 2 He SUBJ 2 0 eats ROOT 3 2 cake OBJ 3 1 cake OBJ (0.3) 3 2 cake OBJ (0.9) VOTED 1 2 He SUBJ 2 0 eats ROOT 3 2 cake OBJ 101