Frame Semantics A Semi-supervised approach to role labeling Summary Semi-supervised Semantic Role Labeling Hagen Fürstenau Department of Computational Linguistics Saarland University (joint work with Mirella Lapata) “FEAST” meeting November 26th, 2008 1 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary Outline Frame Semantics 1 A Semi-supervised approach to role labeling 2 Summary 3 2 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary Frame Semantics Charles J. Fillmore, 1975 & 1981 Definition A frame describes a prototypical situation. It is evoked by a frame evoking element (FEE). It can have several frame elements (roles). 3 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary Frame Semantics Charles J. Fillmore, 1975 & 1981 Definition A frame describes a prototypical situation. It is evoked by a frame evoking element (FEE). It can have several frame elements (roles). Matilde fried the catfish in a heavy iron skillet. 3 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary Frame Semantics Charles J. Fillmore, 1975 & 1981 Definition A frame describes a prototypical situation. It is evoked by a frame evoking element (FEE). It can have several frame elements (roles). Apply_heat Matilde fried the catfish in a heavy iron skillet. FEE 3 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary Frame Semantics Charles J. Fillmore, 1975 & 1981 Definition A frame describes a prototypical situation. It is evoked by a frame evoking element (FEE). It can have several frame elements (roles). Apply_heat Heating_instrument Roles Cook Food Matilde fried the catfish in a heavy iron skillet. FEE 3 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary Frame Semantics Shallow semantic analysis Generalizes well across languages Avoids problem of “universal roles” Can benefit various NLP tasks (IR, QA, ...) 4 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary Frame Semantics Shallow semantic analysis Generalizes well across languages Avoids problem of “universal roles” Can benefit various NLP tasks (IR, QA, ...) 4 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary Frame Semantics Shallow semantic analysis Generalizes well across languages Avoids problem of “universal roles” Can benefit various NLP tasks (IR, QA, ...) 4 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary Frame Semantics Shallow semantic analysis Generalizes well across languages Avoids problem of “universal roles” Can benefit various NLP tasks (IR, QA, ...) How much did Google pay for YouTube? B M Goods u o n y e e y r Commerce_goods-transfer Goods r Money e y u B Google snapped up YouTube for $1.65 billion. 4 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary Frame Semantic Parsing To automatically derive Frame Semantic analyses Take an annotated corpus: FrameNet (135,000 sentences) 1 Train a classifier on this data 2 Use classifier as Frame Semantic parser 3 5 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary Frame Semantic Parsing To automatically derive Frame Semantic analyses Take an annotated corpus: FrameNet (135,000 sentences) 1 Train a classifier on this data 2 Use classifier as Frame Semantic parser 3 Annotation is expensive and time-consuming Must be repeated for new languages or domains Can we reduce this annotation effort? 5 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary Semi-supervised learning Goal: Try to make use of unlabeled data! Example: Binary classification 6 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary Semi-supervised learning Goal: Try to make use of unlabeled data! Example: Binary classification ? 6 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary Semi-supervised learning Goal: Try to make use of unlabeled data! Example: Binary classification ? 6 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary Semi-supervised learning Goal: Try to make use of unlabeled data! Example: Binary classification ? 6 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary Semi-supervised learning Goal: Try to make use of unlabeled data! Example: Binary classification ? 6 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary Semi-supervised learning Goal: Try to make use of unlabeled data! Example: Binary classification ? 6 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary Applied to role labeling To expand a Frame Semantic corpus Find unlabeled sentences “similar” to labeled ones 1 Project annotations from labeled sentences 2 Add new labeled examples to annotation pool 3 Hope that the expanded corpus is “better” 4 than the original one 7 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary Applied to role labeling To expand a Frame Semantic corpus Find unlabeled sentences “similar” to labeled ones 1 Project annotations from labeled sentences 2 Add new labeled examples to annotation pool 3 Hope that the expanded corpus is “better” 4 than the original one What’s “similar”? Take into account syntactic and semantic measures! What’s “better”? A supervised algorithm makes better predictions when trained on the expanded corpus. 7 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary The General Framework FrameNet training test Syntactic parsing ... ... Classifier 8 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary The General Framework FrameNet BNC training test Syntactic Syntactic parsing parsing 8 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary The General Framework FrameNet BNC training test Syntactic Syntactic parsing parsing Annotation of similar sentences 8 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary The General Framework FrameNet BNC training test Classifier Results improved? 8 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary Similarity measure Find best alignment between predicate-argument structures: Apply_heat H e a t i Cook n Food g _ i n fry s boil t r u subj m subj obj m e obj o n d t in Mathilde catfish we egg skillet the some a heavy iron GR Lemma Role GR Lemma Role subj Mathilde Cook subj we obj catfish Food obj egg mod_in skillet Heating_instrument 9 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary Similarity measure Find best alignment between predicate-argument structures: Apply_heat Apply_heat H e a t i Cook n Cook Food Food g _ i n fry s boil t r u subj m subj obj m e obj o n d t in Mathilde catfish we egg skillet the some a heavy iron GR Lemma Role GR Lemma Role subj Mathilde Cook subj we Cook obj catfish Food obj egg Food mod_in skillet Heating_instrument 9 / 17
� � � � � � � � � � � � � � Frame Semantics A Semi-supervised approach to role labeling Summary Alignment We can feel the blood coursing through our veins again. Adrenalin was still coursing through her veins. blood adrenalin Fluid � � � SUBJ SUBJ vein be Path � � � � IOBJ _ THROUGH AUX � � � � � � � � � again still � � � MOD MOD � � � � � � � � vein � � IOBJ _ THROUGH 10 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary Similarity measure Consider a partial, injective alignment function σ : { 1 , . . . , m } → { ε, 1 , . . . , n } ( σ ( i ) = σ ( j ) � = ε ⇒ i = j ) and define similarity with respect to this alignment: m � � � A · δ GR i , GR σ ( i ) + cos ( � v i ,� sim ( σ ) := v σ ( i ) ) − B i = 1 σ ( i ) � = ε The similarity of two predicate-argument structures is sim ( σ ) max σ 11 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary Choosing which sentences to label We have a large corpus, therefore we can be picky: Annotate if similarity is above some threshold? Global threshold value doesn’t work! Pick k-NN unlabeled sentences for each labeled one? Neglects some of the global structure. Global graph optimization? Computationally expensive. 12 / 17
Frame Semantics A Semi-supervised approach to role labeling Summary Choosing which sentences to label We have a large corpus, therefore we can be picky: Annotate if similarity is above some threshold? Global threshold value doesn’t work! Pick k-NN unlabeled sentences for each labeled one? Neglects some of the global structure. Global graph optimization? Computationally expensive. 12 / 17
Recommend
More recommend
