Call Center Conversations Research Articles

Assessing language models’ task and language transfer capabilities for sentiment analysis in dialog data

Our work explores the differences between GRU-based and transformer-based approaches in the context of sentiment analysis on text dialog. In addition to the overall performance on the downstream task, we assess the knowledge transfer capabilities of the models by applying a thorough zero-shot analysis at task level, and on the cross-lingual performance between five European languages. The ability to generalize over different tasks and languages is of high importance, as the data needed for a particular application may be scarce or non existent. We perform evaluations on both known benchmark datasets and a novel synthetic dataset for dialog data, containing Romanian call-center conversations. We study the most appropriate combination of synthetic and real data for fine-tuning on the downstream task, enabling our models to perform in low-resource environments. We leverage the informative power of the conversational context, showing that appending the previous four utterances of the same speaker to the input sequence has the greatest benefit on the inference performance. The cross-lingual and cross-task evaluations have shown that the transformer-based models possess superior transfer abilities to the GRU model, especially in the zero-shot setting. Considering its prior intensive fine-tuning on multiple labeled datasets for various tasks, FLAN-T5 excels in the zero-shot task experiments, obtaining a zero-shot accuracy of 51.27% on the IEMOCAP dataset, alongside the classical BERT that obtained the highest zero-shot accuracy on the MELD dataset with 55.08%.

RESEARCH ON VOICE QUALITY IN CALL CENTER DISCOURSE: VALIDATING RELIABILITY THROUGH INTERRATER AGREEMENT AND PRAAT

Voice quality is an important component of effective communication in call center conversations. A qualitative research methodology was used in the present study to examine how Filipino customer service representatives (CSRs) and American customers use voice quality features in call center discourse. A precise evaluation of voice quality features can aid in the development of effective communication strategies. Through interrater agreement, the present study aimed to validate the reliability of voice quality feature assessment. In this study, how Filipino CSRs and American customers' call center telephone conversations differed in terms of voice quality characteristics such as volume (Loud/Soft), pitch (High/Low), tension (Tense/Lax), and rhythm (Fast/Slow) was investigated. The conversation transcripts were examined using the aforementioned voice quality features. The transcriptions were then evaluated by three independent raters to determine interrater agreement. The findings revealed a high level of interrater agreement between the three raters of all voice quality features, with up to 0.8 in identifying voice quality changes in generic stages. In particular, a higher agreement was found in the assessment of specific voice quality features such as loudness, high pitch, and tension. Praat software was also used to aid in the analysis of some voice recordings to validate the interrater agreement reliability. The approach taken in the present study for assessing voice quality using interrater agreement and Praat software can improve service quality in the call center industry and provides a reliable and valid framework for future qualitative research in applied linguistics.

Author Profiling in Informal and Formal Language Scenarios Via Transfer Learning

The interest in author profiling tasks has increased in the research community because computer applications have shown success in different sectors such as security, marketing, healthcare, and others. Recognition and identification of traits such as gender, age or location based on text data can help to improve different marketing strategies. This type of technology has been widely discussed regarding documents taken from social media. However, its methods have been poorly studied using data with a more formal structure, where there is no access to emoticons, mentions, and other linguistic phenomena that are only present in social media. This paper proposes the use of recurrent and convolutional neural networks and a transfer learning strategy to recognize two demographic traits, i.e., gender and language variety, in documents written in informal and formal language. The models were tested in two different databases consisting of tweets (informal) and call-center conversations (formal). Accuracies of up to 75 % and 68 % were achieved in the recognition of gender in documents with informal and formal language, respectively. Moreover, regarding language variety recognition, accuracies of 92 % and 72 % were obtained in informal and formal text scenarios, respectively. The results indicate that, in relation to the traits considered in this paper, it is possible to transfer the knowledge from a system trained on a specific type of expressions to another one where the structure is completely different and data are scarcer.

INTERACTION ANALYSIS OVER SPEECH FOR CALL CENTRE

Customer Service Center is the second most important consideration just after the actual product. Also, customer service is one of the biggest contributors to the cost component for any firm. We aim to apply well-known data mining techniques to the problem of predicting the quality of interactions like those done in call centers and the problem of predicting the quality of service. The analysis of call center conversations will provide useful insights for enhancing Call Center Analytics to a level that will enable new metrics and key performance indicators (KPIs) beyond the standard approach. These metrics rely on understanding the dynamics of conversations by highlighting the way participants discuss topics. The main focus will be to reduce the average handling time, is a call center metric for the average duration of one transaction, typically measured from the customer’s initiation of the call and including any hold time, talk time and related tasks that follow the transaction. Get real-time solution. The main operations will be speaker diarization, speech to text, agent analysis, emotion recognition and other measures to help with the analysis. We will use RAVDESS (Ryerson Audio-Visual Database of Emotional Speech and Song) for emotion analysis, consisting of vocal emotional expressions in sentences spoken in a range of basic emotional states (happy, sad, anger, fear, disgust, surprise and calm). Emotion recognition is done by extracting features from the audio from its Mel-frequency cepstral coefficients (MFCCs) and passing it through a convolutional neural network. All of this will happen in real time as the call is taking place.

A Fuzzy Approach to Mute Sensitive Information in Noisy Audio Conversations

Audio conversation between a service seeking customer and an agent are common in a voice based call center (VbCC) and are often recorded either for audit purposes or to enable the VbCC to improve their efficiency. These audio recordings invariably containpersonal information of the customer, often spoken by the customer to confirm their identity to get personalized services from the agent. This private to a person (P2aP) information is the recordings is a serious concern from the GDPR perspective and can lead to identity the ftamong other things. In this paper, we propose a robust framework that enables us to reliably spot the P2aP information in the audio and automatically mute it. The main contribution of this paper is the proposal of a novelfuzzy criteria which by design allows for reduced false alarms and at the same time increases the accuracy of the muting process even when the speech to text conversion process is erroneous. Evaluation on real call center conversations demonstrates the reliability of the proposed approach.

Annotating and modeling empathy in spoken conversations

Empathy, as defined in behavioral sciences, expresses the ability of human beings to recognize, understand and react to emotions, attitudes and beliefs of others. In this paper, we address two related problems in automatic affective behavior analysis: the design of the annotation protocol and the automatic recognition of empathy from human–human dyadic spoken conversations. We propose and evaluate an annotation scheme for empathy inspired by the modal model of emotions. The annotation scheme was evaluated on a corpus of real-life, dyadic spoken conversations. In the context of behavioral analysis, we designed an automatic segmentation and classification system for empathy. Given the different speech and language levels of representation where empathy may be communicated, we investigated features derived from the lexical and acoustic spaces. The feature development process was designed to support both the fusion and automatic selection of relevant features from a high dimensional space. The automatic classification system was evaluated on call center conversations where it showed significantly better performance than the baseline.

Automated transcription of conversational Call Center speech – with respect to non-verbal acoustic events

This paper summarizes our recent efforts made to transcribe real-life Call Center conversations automatically with respect to non-verbal acoustic events, as well. Future Call Centers – as cognitive infocom systems – must respond automatically not only for well formed utterances but also for spontaneous and non-word speaker manifestations and must be robust against sudden noises. Conversational telephony speech transcription itself is a big challenge, primarily we address this issue on real-life (Bank and Insurance) tasks. In addition, we introduce several non-word acoustic modeling approaches and their integration to LVCSR (Large Vocabulary Continuous Speech Recognition). In the experiments, one and two channel (client and agent speech merged into one or left in two separate audio stream) transcription results, cross-task results and the handling of transcription data insufficiency are investigated – in parallel with the non-verbal acoustic event modeling. On the agent side less than 15% word error rate could be achieved and the best error rate reduction is 20% (relative) due to the inclusion of various written corpora and due to acoustic event handling.

The A mitié s system: Data-driven techniques for automated dialogue

We present a natural-language customer service application for a telephone banking call center, developed as part of the A mitié s dialogue project (Automated Multilingual Interaction with Information and Services). Our dialogue system, based on empirical data gathered from real call-center conversations, features data-driven techniques that allow for spoken language understanding despite speech recognition errors, as well as mixed system/customer initiative and spontaneous conversation. These techniques include robust named-entity extraction, slot-filling Frame Agents, vector-based task identification and dialogue act classification, a Bayesian database record selection algorithm, and a natural language generator designed with templates created from real agents’ expressions. Preliminary evaluation results indicate efficient dialogues and high user satisfaction, with performance comparable to or better than that of current conversational information systems.

A Language Independent Approach to Identify Problematic Conversations in Call Centers

Voice based call centers enable customers query for information by speaking to human agents. Most often these call conversations are recorded by call centers with the intent of trying to identify things that can help improve the performance of the call center to serve the customer better. Today the recorded conversations are analyzed by humans by listening to call conversations, which is both time consuming, fatigue prone and not very accurate. Additionally, humans are able to analyze only a small percentage of the total calls because of economics. In this paper which is based on [1], we propose a visual method to identify problem calls quickly. The idea is to sieve through all the calls and identify problem calls, these calls can then be further analyzed by human. We first model call conversations as a directed graph and then identify a directed graph structure associated with a normal call. All call conversations that do not have the structure of a normal call are then classified as being abnormal. We use the speaking rate feature to model call conversation because it can spot potential problem calls. We have experimented on real call center conversations acquired from different call centers and the results are encouraging.