Abstract

In recent years, text-to-speech (TTS) systems have shown considerable improvement as far as the quality of the synthetic speech is concerned. Data driven synthesis methods using syllable as basic unit for concatenation, have proved to generate high quality speech for Indian Languages because of their advantage of prosodic matching function. However, still there is no acceptable solution to the optimal selection of speech segments in terms of audible discontinuities and human perception. This problem gets aggravated in the cases where there is no enough data for building the voice due to the missing units. In this paper, we continue our efforts in trying to address this by investigating the use of a new continuity measure based on maximum signal correlation for optimal selection of units in concatenative text-to-speech (TTS) synthesis framework. We explore two formulations for calculating the signal correlation: cross correlation (CC) based and average magnitude difference function (AMDF) based. We first perform an initial experiment to understand the significance of the approach and then build 5 experimental systems. Evaluations on 30 sentences for each of these languages by native users of the language show that the proposed continuity measure results in more natural sounding synthesis.

Abstract

The present study focuses on the text-independent speaker recognition in emotional conditions. In this paper, both system and source features are considered to represent speaker specific information. At the model level, Gaussian Mixture Models (GMMs), Gaussian Mixture Model-Universal Background Model (GMM-UBM) and Deep Neural Networks (DNN) are explored. The experiments are performed using 3 emotional databases, i.e. German emotional speech database (EMO-DB), IITKGP-SESC: Hindi and IITKGP-SESC: Telugu databases. The emotions considered in the present study are neutral, anger, happy and sad. The results show that, the performance of a speaker recognition system trained with clean speech is degrading while testing with emotional data irrespective of feature used or model used to build the system. The best results are obtained for the score level fusion of system and source features based systems when speakers are modeled with DNNs.

Abstract

Due to increase of online product reviews posted daily through various modalities such as video, audio and text, sentimental analysis has gained huge attention. Recent developments in web technologies have also enabled the increase of web content in Hindi. In this paper, an approach to detect the sentiment of an online Hindi product reviews based on its multi-modality natures (audio and text) is presented. For each audio input, Mel Frequency Cepstral Coefficients (MFCC) features are extracted. These features are used to develop a sentiment models using Gaussian Mixture Models (GMM) and Deep Neural Network (DNN) classifiers. From results, it is observed that DNN classifier gives better results compare to GMM. Further textual features are extracted from the transcript of the audio input by using Doc2vec vectors. Support Vector Machine (SVM) classifier is used to develop a sentiment model using these textual features. From experimental results it is observed that combining both the audio and text features results in improvement in the performance for detecting the sentiment of an online product reviews.

Abstract

Internet users today prefer getting precise answers to their questions rather than sifting through a bunch of relevant documents provided by search engines. This has led to the huge popularity of Community Question Answering (cQA) services like Yahoo! Answers, Baidu Zhidao, Quora, Stack Over flowetc., where forum users respond to questions with precise answers. Over time, such cQA archives become rich repositories of knowledge encoded in the form of questions and user generated answers. In cQA archives, retrieval of similar questions, which have already been answered in some form, is important for improving the effectiveness of such forums. The main challenge while retrieving similar questions is the “lexico-syntactic” gap between the user query and the questions already present in the forum. In this paper, we pro-pose a novel approach called “Deep Structured Topic Model (DSTM)” to bridge the lexico-syntactic gap between the question posed by the user and forum questions. DSTM employs a two-step process consisting of initially retrieving similar questions that lie in the vicinity of the query and latent topic vector space and then re-ranking them using a deep layered semantic model. Experiments on large scale real-life cQA dataset show that our approach outperforms the state-of-the-art translation and topic based baseline approaches

Abstract

It has been shown that transition-based methods can be used for syntactic word ordering and tree linearization, achieving significantly faster speed compared with traditional best-first methods. State-of-the-art transition-based models give competitive results on ab-stract word ordering and unlabeled tree lin-earization, but significantly worse results on labeled tree linearization. We demonstrate that the main cause for the performance bottle-neck is the sparsity of SHIFT transition actions rather than heavy pruning. To address this is-sue, we propose a modification to the standard transition-based feature structure, which reduces feature sparsity and allows look ahead features at a small cost to decoding efficiency.Our model gives the best reported accuracies on all benchmarks, yet still being over 30times faster compared with best-first-search.

Abstract

Community Question Answering (cQA) forums have become a popular medium for soliciting direct answers to specific questions of users from experts or other experienced users on a given topic. However, for a given question, users some-times have to sift through a large number of low-quality or irrelevant answers to find out the answer which satisfies their information need. To alleviate this, the problem of Answer Quality Prediction (AQP) aims to predict the quality of an answer posted in response to a forum question.Current AQP systems either learn models using - a) various hand-crafted features (HCF) or b) use deep learning (DL)techniques which automatically learn the required feature representations.In this paper, we propose a novel approach for AQP known as -“Deep Feature Fusion Network (DFFN)”which leverages the advantages of both hand-crafted features and deep learning based systems. Given a question-answer pair along with its metadata, DFFN independently - a) learns deep features using a Convolutional Neural Network (CNN) and b) computes hand-crafted features using various external resourcesand then combines them using a deep neural network trained to predict the final answer quality. DFFN achieves state-of-the-art performance on the standard SemEval-2015 andSemEval-2016 benchmark datasets and outperforms base-line approaches which individually employ either HCF or DL based techniques alone.

Abstract

Community Question Answering (cQA)services like Yahoo!Answers1, BaiduZhidao2, Quora3, StackOverflow4etc.provide a platform for interaction with experts and help users to obtain precise and accurate answers to their questions.The time lag between the user posting a question and receiving its answer could be reduced by retrieving similar historic questions from the cQA archives.The main challenge in this task is the “lexico-syntactic” gap between the current and the previous questions. In this paper, we pro-pose a novel approach called“Siamese Convolutional Neural Network for cQA(SCQA)”to find the semantic similarity between the current and the archived ques-tions.SCQA consist of twin convolutional neural networks with shared parameters and a contrastive loss function joining them. SCQA learns the similarity metric for question-question pairs by leveraging the question-answer pairs available in cQA forum archives. The model projects semantically similar question pairs nearer to each other and dissimilar question pairs farther away from each other in the semantic space. Experiments on large scale real-life “Yahoo! Answers” dataset reveals that SCQA outperforms current state-of-the-art approaches based on translation models, topic models and deep neural network based models which use non-shared parameters.

Abstract

In this paper we describe an end to end Neural Model for Named Entity Recognition (NER) which is based on Bi-Directional RNN-LSTM. Almost all NER systems for Hindi use Language Specific features and handcrafted rules with gazetteers. Our model is language independent and uses no domain specific features or any handcrafted rules. Our models rely on semantic information in the form of word vectors which are learnt by an unsupervised learning algorithm on an unannotated corpus. Our model attained state of the art performance in both English and Hindi without the use of any morphological analysis or without using gazetteers of any sort.

Abstract

Community Question Answering (cQA) forums have become a popular medium for soliciting answers to specific user questions from experts and experienced users in a given topic. However,for a given question, users sometimes have to sift through a large number of low-quality or irrelevant answers to find out the answer which satisfies their information need. To alleviate this,the problem of Answer Quality Prediction (AQP) aims to predict the quality of an answer posted in response to a forum question. Current AQP systems either learn models using - a) various hand-crafted features (HCF) or b) Deep Learning (DL) techniques which automatically learn the feature representations.In this paper, we propose a novel approach for AQP known as -“Deep Feature Fusion Network(DFFN)”which combines the advantages of both hand-crafted features and deep learning based systems. Given a question-answer pair along with its metadata, a DFFN architecture indepen-dently - a) learns features using the Deep Neural Network (DNN) and b) computes handcrafted features leveraging various external resources and then combines the musing a fully connected neural network trained to predict the quality of the given answer. DFFN is an end-end differen-tiable model and trained as a single system. We propose two different DFFN architectures which vary mainly in the way they model the input question/answer pair - a) DFFN-CNN which uses a Convolutional Neural Network (CNN) and b) DFFN-BLNA which uses a Bi-directional LSTMwith Neural Attention (BLNA). Both these proposed variants of DFFN (DFFN-CNN and DFFN-BLNA) achieve state-of-the-art performance on the standard SemEval-2015 and SemEval-2016benchmark datasets and outperforms baseline approaches which individually employ either HCFor DL based techniques alone.

Abstract

Today, a large amount of audio data is available on the web in the form of audiobooks, podcasts, video lectures, video blogs, news bulletins, etc. In addition, we can effortlessly record and store audio data such as a read, lecture, or impromptu speech on handheld devices. These data are rich in prosody and provide a plethora of voices to choose from, and their availability can significantly reduce the overhead of data preparation and help rapid building of synthetic voices. But, a few problems are associated with readily using this data such as (1) these audio files are generally long, and audio-transcription alignment is memory intensive; (2) precise corresponding transcriptions are unavailable, (3) many times, no transcriptions are available at all; (4) the audio may contain dis-fluencies and non-speech noises, since they are not specifically recorded for building synthetic voices; and (5) if we obtain automatic transcripts, they will not be error free. Earlier works on long audio alignment addressing the first and second issue generally preferred reasonable transcripts and mainly focused on (1) less manual intervention, (2) mispronunciation detection, and (3) segmentation error recovery. In this work, we use a large vocabulary public domain automatic speech recognition (ASR) system to obtain transcripts, followed by confidence measure-based data pruning which together address the five issues with the found data and also ensure the above three points. For proof of concept, we build voices in the English language using an audiobook (read speech) in a female voice from LibriVox and a lecture (spontaneous speech) in a male voice from Coursera, using both reference and hypotheses transcriptions, and evaluate them in terms of intelligibility and naturalness with the help of a perceptual listening test on the Blizzard 2013 corpus.