How to run HAABSADA, a Hybrid Approach for Aspect Based Sentiment Analysis with Data Augmentation

Data Augmentation extenstion on HAABSA. In order to use the code please follow the following steps:

1. Go to https://github.com/ofwallaart/HAABSA and follow the steps on the github page. There are some possible problems that you can come across when following his, which I hope to tackle as much as possible.

• When running on windows, note that you need compatible c++ visual studio build tools (https://visualstudio.microsoft.com/downloads/).
• If you want to run on GPU (NVIDIA) note that you need compatible drivers, Cuda (v9.0) and cuDNN (v6.4)
• If you need nltk files --> in your environment run python afterwards import nltk afterwards ntlk.download(NLTK_FILE)
• Make sure your environment is active (with Scripts/activate) --> you should see the name of your environment next to the command line

2. Clone this repository using git clone https://github.com/tomasLiesting/HAABSADA.git and, in your file explorer, overwrite the files in HAABSA with the files in this repo
3. In your environment, cd to HAABSA, and again run pip install -r requirements.txt
4. Make sure you have access to the google translate api, if you want to use backtranslation. You have to create an account and get 300€ for free. This is done as follows:

• pip install --upgrade google-cloud-translate
• Afterwards go to this page https://cloud.google.com/docs/authentication/getting-started#windows
• Create a service account
• Create a new project
• Create a service account with as role project owner
• Download the json authentication (and place it somewhere you like)
• In the environment run $env:GOOGLE_APPLICATION_CREDENTIALS="[PATH]" (where PATH is the path to your json authentication, e.g. $env:GOOGLE_APPLICATION_CREDENTIALS="C:\Users\username\Downloads\my-key.json"

5. In the environment, run python main.py

How to modify parameters

In the file config.py, there are several Data Augmentation variables. These can be modified. To reproduce the 1% and 0.5% improvement on the 2015 and 2016 SemEval datasets, make sure that:

• EDA_type = 'adjusted'
• EDA_deletion = 0
• EDA_replacement = 1
• EDA_insertion = 1
• EDA_swap = 1
• EDA_pct = .2
• backtranslation_langs = 'None'
• use_word_mixup = 0
• original_multiplier = 3

Furthermore, in main.py make sure that:

• loadData = True
• augment_data = True
• useOntology = True
• runLCRROTALT = True For the rest all parameters in main.py should be False.