Discriminative spoken language understanding using word confusion networks

Current commercial dialogue systems typically use hand-crafted grammars for Spoken Language Understanding (SLU) operating on the top one or two hypotheses output by the speech recogniser. These systems are expensive to develop and they suffer from significant degradation in performance when faced with recognition errors. This paper presents a robust method for SLU based on features extracted from the full posterior distribution of recognition hypotheses encoded in the form of word confusion networks. Following [1], the system uses SVM classifiers operating on n-gram features, trained on unaligned input/output pairs. Performance is evaluated on both an off-line corpus and on-line in a live user trial. It is shown that a statistical discriminative approach to SLU operating on the full posterior ASR output distribution can substantially improve performance both in terms of accuracy and overall dialogue reward. Furthermore, additional gains can be obtained by incorporating features from the previous system output. © 2012 IEEE.

Cross-Lingual spoken language understanding from unaligned data using discriminative classification models and machine translation

This paper investigates several approaches to bootstrapping a new spoken language understanding (SLU) component in a target language given a large dataset of semantically-annotated utterances in some other source language. The aim is to reduce the cost associated with porting a spoken dialogue system from one language to another by minimising the amount of data required in the target language. Since word-level semantic annotations are costly, Semantic Tuple Classifiers (STCs) are used in conjunction with statistical machine translation models both of which are trained from unaligned data to further reduce development time. The paper presents experiments in which a French SLU component in the tourist information domain is bootstrapped from English data. Results show that training STCs on automatically translated data produced the best performance for predicting the utterance's dialogue act type, however individual slot/value pairs are best predicted by training STCs on the source language and using them to decode translated utterances. © 2010 ISCA.

Syllable language models for Mandarin speech recognition: exploiting character language models.

Mandarin Chinese is based on characters which are syllabic in nature and morphological in meaning. All spoken languages have syllabiotactic rules which govern the construction of syllables and their allowed sequences. These constraints are not as restrictive as those learned from word sequences, but they can provide additional useful linguistic information. Hence, it is possible to improve speech recognition performance by appropriately combining these two types of constraints. For the Chinese language considered in this paper, character level language models (LMs) can be used as a first level approximation to allowed syllable sequences. To test this idea, word and character level n-gram LMs were trained on 2.8 billion words (equivalent to 4.3 billion characters) of texts from a wide collection of text sources. Both hypothesis and model based combination techniques were investigated to combine word and character level LMs. Significant character error rate reductions up to 7.3% relative were obtained on a state-of-the-art Mandarin Chinese broadcast audio recognition task using an adapted history dependent multi-level LM that performs a log-linearly combination of character and word level LMs. This supports the hypothesis that character or syllable sequence models are useful for improving Mandarin speech recognition performance.

Modified Kanerva model. Results for real time word recognition

This paper describes results obtained using the modified Kanerva model to perform word recognition in continuous speech after being trained on the multi-speaker Alvey 'Hotel' speech corpus. Theoretical discoveries have recently enabled us to increase the speed of execution of part of the model by two orders of magnitude over that previously reported by Prager & Fallside. The memory required for the operation of the model has been similarly reduced. The recognition accuracy reaches 95% without syntactic constraints when tested on different data from seven trained speakers. Real time simulation of a model with 9,734 active units is now possible in both training and recognition modes using the Alvey PARSIFAL transputer array. The modified Kanerva model is a static network consisting of a fixed nonlinear mapping (location matching) followed by a single layer of conventional adaptive links. A section of preprocessed speech is transformed by the non-linear mapping to a high dimensional representation. From this intermediate representation a simple linear mapping is able to perform complex pattern discrimination to form the output, indicating the nature of the speech features present in the input window.

Spoken document representations for probabilistic retrieval

This paper presents some developments in query expansion and document representation of our spoken document retrieval system and shows how various retrieval techniques affect performance for different sets of transcriptions derived from a common speech source. Modifications of the document representation are used, which combine several techniques for query expansion, knowledge-based on one hand and statistics-based on the other. Taken together, these techniques can improve Average Precision by over 19% relative to a system similar to that which we presented at TREC-7. These new experiments have also confirmed that the degradation of Average Precision due to a word error rate (WER) of 25% is quite small (3.7% relative) and can be reduced to almost zero (0.2% relative). The overall improvement of the retrieval system can also be observed for seven different sets of transcriptions from different recognition engines with a WER ranging from 24.8% to 61.5%. We hope to repeat these experiments when larger document collections become available, in order to evaluate the scalability of these techniques.

KM your way to CMMI

Purpose - The purpose of this paper is to describe two related fields - knowledge management (KM) and capability maturity model integrated (CMMISM) and highlight their imilarities. Design/methodology/approach - The KM framework used for this comparison is the one established and used at Israel Aircraft Industries, while the CMMISM source of information is none but the original document produced by the CMMISM product team at the Carnegie Mellon University, as well as papers published on the subject. Findings - Knowledge management is a rather young discipline promising to maximize innovation and competitive advantage to organizations that practice knowledge capture, documentation, retrieval and reuse, creation, transfer and share to its knowledge assets in a measurable way, integrated in its operational and business processes. The capability maturity model integrated deals with the ways an organization has to follow, in order to maintain well mapped processes, having well defined stages, because of the assumption that in mature organizations, it is possible to measure and relate between the quality of the process and the quality of the product. Though KM and CMMISM take different approaches to the achievement of competitive advantage, they seem to be supporting as well as dependent of each other. Originality/value - Practitioners as well as researchers in the field of knowledge management and in the implementation of the CMMISM standard will find comfort in realizing how mutually supportive are these two fields. © Emerald Group Publishing Limited.

Differentiated adipose-derived stem cells act synergistically with RGD-modified surfaces to improve neurite outgrowth in a co-culture model.

Peripheral nerve damage is a problem encountered after trauma and during surgery and the development of synthetic polymer conduits may offer a promising alternative to autografts. In order to improve the performance of the polymer to be used for nerve conduits, poly-ε-caprolactone (PCL) films were chemically functionalized with RGD moieties, using a chemical reaction previously developed. In vitro cultures of dissociated dorsal root ganglion (DRG) neurons provide a valid model to study different factors affecting axonal growth. In this work, DRG neurons were cultured on RGD-functionalized PCL films. Adult adipose-derived stem cells differentiated to Schwann cells (dASCs) were initially cultured on the functionalized PCL films, resulting in improved attachment and proliferation. dASCs were also co-cultured with DRG neurons on treated and untreated PCL to assess stimulation by dASCs on neurite outgrowth. Neuron response was generally poor on untreated PCL films, but long neurites were observed in the presence of dASCs or RGD moieties. A combination of the two factors enhanced even further neurite outgrowth, acting synergistically. Finally, in order to better understand the extracellular matrix (ECM)-cell interaction, a β1 integrin blocking experiment was carried out. Neurite outgrowth was not affected by the specific antibody blocking, showing that β1 integrin function can be compensated by other molecules present on the cell membrane. Copyright © 2013 John Wiley & Sons, Ltd.

Aeroacoustic source measurement methods for characterizing the sound generated by ducted flow devices with higher-order modes.

The International Organization for Standardization (ISO) method 5136 is widely used in industry and academia to determine the sound power radiated into a duct by fans and other flow devices. The method involves placing the device at the center of a long cylindrical duct with anechoic terminations at each end to eliminate reflections. A single off-axis microphone is used on the inlet and outlet sides that can theoretically capture the plane-wave mode amplitudes but this does not provide enough information to fully account for higher-order modes. In this study, the "two-port" source model is formulated to include higher-order modes and applied for the first three modes. This requires six independent surface pressure measurements on each side or "port." The resulting experimental set-up is much shorter than the ISO rig and does not require anechoic terminations. An array of six external loudspeaker sources is used to characterize the passive part of the two-port model and the set-up provides a framework to account for transmission of higher-order modes through a fan. The relative importance of the higher-order modes has been considered and their effect on inaccuracies when using the ISO method to find source sound power has been analyzed.