Advertisement

Integration of Word and Semantic Features for Theme Identification in Telephone Conversations

  • Yannick Estève
  • Mohamed Bouallegue
  • Carole Lailler
  • Mohamed Morchid
  • Richard Dufour
  • Georges Linarès
  • Driss Matrouf
  • Renato De Mori

Abstract

The paper describes a research about the possibility of integrating different types of word and semantic features for automatically identifying themes of real-life telephone conversations in a customer care service (CCS). Features are all the words of the application vocabulary, the probabilities obtained with latent Dirichlet allocation (LDA) of selected discriminative words and semantic features obtained with a limited human supervision of words and patterns expressing entities and relations of the application ontology. A deep neural network (DNN) is proposed for integrating these features. Experimental results on manual and automatic conversation transcriptions are presented showing the effective contribution of the integration. The results show how to automatically select a large subset of the test corpus with high precision and recall, making it possible to automatically obtain theme mention proportions in different time periods.

Keywords

Theme identification Human–human spoken conversation Deep neural network 

References

  1. Béchet F, Maza B, Bigouroux N, Bazillon T, El-Bèze M, De Mori R, Arbillot E (2012) Decoda: a call-centre human-human spoken conversation corpus. In: Proceeding of LREC’12Google Scholar
  2. Blei DM, Ng AY, Jordan MI (2003) Latent Dirichlet allocation. J Mach Learn Res 3:993–1022MATHGoogle Scholar
  3. Carpineto C, De Mori R, Romano G, Bigi B (2001) An information-theoretic approach to automatic query expansion. ACM Trans Inf Syst 19(1):1–27CrossRefGoogle Scholar
  4. Chen Y-N, Wang WY, Rudnicky AI (2014) Leveraging frame semantics and distributional semantics for unsupervised semantic slot induction in spoken dialogue systems. In: IEEE spoken language technology workshop (SLT 2014), South Lake Tahoe, California and NevadaGoogle Scholar
  5. Cuayáhuitl H, Dethlefs N, Hastie H, Liu X (2014) Training a statistical surface realiser from automatic slot labelling. In: IEEE spoken language technology workshop (SLT 2014), South Lake Tahoe, California and NevadaGoogle Scholar
  6. Deerwester S, Dumais ST, Furnas GW, Landauer TK, Harshman R (1990) Indexing by latent semantic analysis. J Am Soc Inf Sci 41(6):391–407CrossRefGoogle Scholar
  7. Hazen TJ (2011) MCE training techniques for topic identification of spoken audio documents. IEEE Trans Audio Speech Lang Process 19(8):2451–2460CrossRefGoogle Scholar
  8. Linarès G, Nocéra P, Massonie D, Matrouf D (2007) The LIA speech recognition system: from 10xRT to 1xRT. In: Proceedings of the 10th international conference on text, speech and dialogue. Springer, Berlin, pp 302–308Google Scholar
  9. Morchid M, Dufour R, Bousquet P-M, Bouallegue M, Linarès G, De Mori R (2014) Improving dialogue classification using a topic space representation and a gaussian classifier based on the decision rule. In: Proceedings of ICASSPCrossRefGoogle Scholar
  10. Morchid M, Bouallegue M, Dufour R, Linarès G, Matrouf D, De Mori R (2014) An i-vector based approach to compact multi-granularity topic spaces representation of textual documents. In: The 2014 conference on empirical methods on natural language processing (EMNLP), SIGDATGoogle Scholar
  11. Morchid M, Bouallegue M, Dufour R, Linarès G, Matrouf D, De Mori R (2014) I-vector based representation of highly imperfect automatic transcriptions. In: Conference of the international speech communication association (INTERSPEECH) 2014, ISCAGoogle Scholar
  12. Morchid M, Dufour R, Bouallegue M, Linarès G, De Mori R (2014) Theme identification in human-human conversations with features from specific speaker type hidden spaces. In: Fifteenth annual conference of the international speech communication associationGoogle Scholar
  13. Sarikaya R, Hinton GE, Deoras A (2014) Application of deep belief networks for natural language understanding. IEEE/ACM Trans Audio Speech Lang Process 22(4):778–784CrossRefGoogle Scholar
  14. Tur G, De Mori R (2011) Spoken language understanding: systems for extracting semantic information from speech. Wiley, New YorkCrossRefMATHGoogle Scholar
  15. Tur G, Hakkani-Tür D (2011) Human/human conversation understanding. In: Spoken language understanding: systems for extracting semantic information from speech. Wiley, New York, pp 225–255Google Scholar
  16. Wu MS, Lee HS, Wang HM (2010) Exploiting semantic associative information in topic modeling. In: Proceedings of the IEEE workshop on spoken language technology (SLT 2010), pp 384–388Google Scholar

Copyright information

© Springer International Publishing Switzerland 2015

Authors and Affiliations

  • Yannick Estève
    • 1
  • Mohamed Bouallegue
    • 1
  • Carole Lailler
    • 1
  • Mohamed Morchid
    • 2
  • Richard Dufour
    • 2
  • Georges Linarès
    • 2
  • Driss Matrouf
    • 2
  • Renato De Mori
    • 2
    • 3
  1. 1.LIUMUniversity of Le MansLe MansFrance
  2. 2.LIAUniversity of AvignonAvignonFrance
  3. 3.McGill UniversityMontrealCanada

Personalised recommendations