The majority of work in targeted sentiment analysis has concentrated on finding better methods to improve the overall results. Within this paper we show that these models are not robust to linguistic phenomena, specifically negation and speculation. In this paper, we propose a multi-task learning method to incorporate information from syntactic and semantic auxiliary tasks, including negation and speculation scope detection, to create English-language models that are more robust to these phenomena. Further we create two challenge datasets to evaluate model performance on negated and speculative samples. We find that multi-task models and transfer learning via language modelling can improve performance on these challenge datasets, but the overall performances indicate that there is still much room for improvement. We release both the datasets and the source code at <a href=”https://github.com/jerbarnes/multitask_negation_for_targeted_sentiment”>https://github.com/jerbarnes/multitask_negation_for_targeted_sentiment</a>.

We present FIESTA, a model selection approach that significantly reduces the computational resources required to reliably identify state-of-the-art performance from large collections of candidate models. Despite being known to produce unreliable comparisons, it is still common practice to compare model evaluations based on single choices of random seeds. We show that reliable model selection also requires evaluations based on multiple train-test splits (contrary to common practice in many shared tasks). Using bandit theory from the statistics literature, we are able to adaptively determine appropriate numbers of data splits and random seeds used to evaluate each model, focusing computational resources on the evaluation of promising models whilst avoiding wasting evaluations on models with lower performance. Furthermore, our user-friendly Python implementation produces confidence guarantees of correctly selecting the optimal model. We evaluate our algorithms by selecting between 8 target-dependent sentiment analysis methods using dramatically fewer model evaluations than current model selection approaches.

Lack of repeatability and generalisability are two significant threats to continuing scientific development in Natural Language Processing. Language models and learning methods are so complex that scientific conference papers no longer contain enough space for the technical depth required for replication or reproduction. Taking Target Dependent Sentiment Analysis as a case study, we show how recent work in the field has not consistently released code, or described settings for learning methods in enough detail, and lacks comparability and generalisability in train, test or validation data. To investigate generalisability and to enable state of the art comparative evaluations, we carry out the first reproduction studies of three groups of complementary methods and perform the first large-scale mass evaluation on six different English datasets. Reflecting on our experiences, we recommend that future replication or reproduction experiments should always consider a variety of datasets alongside documenting and releasing their methods and published code in order to minimise the barriers to both repeatability and generalisability. We have released our code with a model zoo on GitHub with Jupyter Notebooks to aid understanding and full documentation, and we recommend that others do the same with their papers at submission time through an anonymised GitHub account.

This paper describes our participation in Task 5 track 2 of SemEval 2017 to predict the sentiment of financial news headlines for a specific company on a continuous scale between -1 and 1. We tackled the problem using a number of approaches, utilising a Support Vector Regression (SVR) and a Bidirectional Long Short-Term Memory (BLSTM). We found an improvement of 4-6% using the LSTM model over the SVR and came fourth in the track. We report a number of different evaluations using a finance specific word embedding model and reflect on the effects of using different evaluation metrics.

Attribution bias refers to the tendency of people to attribute successes to their own abilities but failures to external factors. In a business context an internal factor might be the restructuring of the firm and an external factor might be an unfavourable change in exchange or interest rates. In accounting research, the presence of an attribution bias has been demonstrated for the narrative sections of the annual financial reports. Previous studies have applied manual content analysis to this problem but in this paper we present novel work to automate the analysis of attribution bias through using machine learning algorithms. Previous studies have only applied manual content analysis on a small scale to reveal such a bias in the narrative section of annual financial reports. In our work a group of experts in accounting and finance labelled and annotated a list of 32,449 sentences from a random sample of UK Preliminary Earning Announcements (PEAs) to allow us to examine whether sentences in PEAs contain internal or external attribution and which kinds of attributions are linked to positive or negative performance. We wished to examine whether human annotators could agree on coding this difficult task and whether Machine Learning (ML) could be applied reliably to replicate the coding process on a much larger scale. Our best machine learning algorithm correctly classified performance sentences with 70% accuracy and detected tone and attribution in financial PEAs with accuracy of 79%.