초록 ▼

Various embodiments are directed to a system and method providing associative matching of terms. Candidate terms are selected for building one or more associative matching models from one or more selected candidate sources. Associativity is defined to give editors the ability to label sample associa

Various embodiments are directed to a system and method providing associative matching of terms. Candidate terms are selected for building one or more associative matching models from one or more selected candidate sources. Associativity is defined to give editors the ability to label sample associative term pairs from the one or more candidate sources. The editors label sample candidate term pairs as being related. Features are determined that can differentiate associative from non-associative pairs. The selected features are used to build a model. The model is applied to determine whether a received query-candidate pair are associative.

대표청구항 ▼

What is claimed is: 1. A computerized method for identification of term or phrase pairs that are associative matches, the method comprising: electronically selecting one or more candidate term pairs for the construction of an associative model; electronically labeling the one or more candidate term

What is claimed is: 1. A computerized method for identification of term or phrase pairs that are associative matches, the method comprising: electronically selecting one or more candidate term pairs for the construction of an associative model; electronically labeling the one or more candidate term pairs by matching the one or more candidate terms on the basis of associative criteria, the associative criteria comprising a complementary product relationship, a task relationship, a cultural relationship and temporal relationships; electronically determining one or more features for the one or more candidate term pairs; electronically building a model on the basis of the one or more features; and electronically applying the model to an unlabeled candidate term pair to determine if the pair is an associative match. 2. The method of claim 1 comprising selecting one or more sources from which the one or more candidate terms may be selected. 3. The method of claim 2 wherein the one or more sources are selected from the set of sources that include a database of advertising bids, query logs from user queries, social media, on-line communities and scanner data from receipts. 4. The method of claim 1 comprising identifying the one or more associative criteria. 5. The method of claim 1 wherein determining a given feature comprises determining co-bidedness, brand relationship and statistical features. 6. The method of claim 1 wherein building the model comprises building a first sub-model and a second sub-model. 7. The method of claim 6 wherein building the first sub-model comprises building a model to predict whether the terms of a given candidate term pair have a semantic relationship. 8. The method of claim 6 wherein building the second sub-model comprises building a model to predict whether or not the candidate term pair have a broad relationship. 9. The method of claim 6 wherein the first sub-model and the second sub-model predict specific aggregate re-combinations of one or more labels. 10. The method of claim 6 wherein the first sub-model and the second sub-model are selected from the set of models that include a logistic regression, a neural network, a decision tree and a support vector machine. 11. The method of claim 1 wherein applying the model occurs in response to a search query. 12. The method of claim 1 comprising generating a data store to maintain one or more associative matches. 13. The method of claim 12 comprising consulting the data store upon receipt of one or more query terms to determine one or more terms that are associative matches with a given one of the one or more query terms. 14. The method of claim 1 wherein the model is selected from the set of models that include a logistic regression, a neural network, a decision tree and a support vector machine. 15. Computer readable media comprising program code that when executed by a programmable processor causes the processor to execution a method for identification of term or phrase pairs that are associative matches, the computer readable media comprising: program code stored on physical computer readable media for selecting one or more candidate term pairs for the construction of an associative model; program code stored on physical computer readable media for labeling the one or more candidate term pairs by matching the one or more candidate terms on the basis of associative criteria, the associative criteria comprising a complementary product relationship, a task relationship, a cultural relationship and temporal relationships; program code stored on physical computer readable media for determining one or more features for the one or more candidate term pairs; program code stored on physical computer readable media for building a model on the basis of the one or more features; and program code stored on physical computer readable media for applying the model to an unlabeled candidate term pair to determine if the pair is an associative match. 16. The computer readable media of claim 15 comprising program code for selecting one or more sources from which the one or more candidate terms may be selected. 17. The computer readable media of claim 16 wherein the one or more sources are selected from the set of sources that include a database of advertising bids, query logs from user queries, social media, on-line communities and scanner data from receipts. 18. The computer readable media of claim 15 comprising program code for identifying the one or more associative criteria. 19. The computer readable media of claim 15 wherein program code for determining a given feature comprises program code for determining co-bidedness, brand relationship and statistical features. 20. The computer readable media of claim 15 wherein program code for building the model comprises program code for building a first sub-model and a second sub-model. 21. The computer readable media of claim 20 wherein program code for building the first sub-model comprises program code for building a model to predict whether the terms of a given candidate term pair have a semantic relationship. 22. The computer readable media of claim 20 wherein program code for building the second sub-model comprises program code for building a model to predict whether or not the candidate term pair have a broad relationship. 23. The computer readable media of claim 20 wherein the first sub-model and the second sub-model predict specific aggregate re-combinations of one or more labels. 24. The computer readable media of claim 20 wherein the first sub-model and the second sub-model are selected from the set of models that include a logistic regression, a neural network, a decision tree and a support vector machine. 25. The computer readable media of claim 20 wherein program code for applying the model is executed by the processor in response to a search query. 26. The computer readable media of claim 15 comprising program code for consulting the data store upon receipt of one or more query terms to determine one or more terms that are associative matches with a given one of the one or more query terms. 27. The computer readable media of claim 15 comprising program code for generating a data store to maintain one or more associative matches. 28. The computer readable media of claim 15 wherein program code for building a model is selected form the set of models that include a logistic regression, a neural network, a decision tree and a support vector machine.