DV victims are always isolated. During the unprecedented Covid-19 lockdown and social distancing conditions at present, the situation is much worse. Being in lockdown is difficult for everyone - imagine for a moment how it must feel to be cut off from everyone except your abuser - the person who hurts you, controls you, who makes you feel bad about yourself... No other friends, no extended family, no colleagues at work, no kind strangers at the shops, no human contact at all... other than the one who hurts you! The dire nature of the situation at the moment is reflected in the statistics published by refuge, the main UK domestic violence charity and reported widely in the media - a 120% increase in calls to their helpline and a 700% increase in website enquiries. It has even been reflected in the government policy for lockdown in the UK - stating that escape from domestic violence is a legitimate reason to leave your home!
What it does
It appears to be a news app, but has hidden functionality. The main component of this is a chatbot which can recognise early warning signs of many different kinds of abuse and recommend the user keep an eye out for this. However, it can also track what level of immediate danger the user might be in and recommend what kind of action the user should take, for example whether they should confide in a friend or call emergency services. It has an understanding of many kinds of abusers and their methods, for children and teens experiencing child abuse or adults experiencing domestic abuse.
A way of providing access to information, advice, support, help and services in a way that is easily accessed, available 24/7, sensitive, secure and private. A one-stop app which can tie together access to information and support services such as DV charities, emergency services, medical care, a safe and secure social network, local resources, mental health advice and services, financial and legal information. Using smart AI and ML tech to create a chat bot interface which makes the user experience straight forward, intuitive, friendly and sensitive. Also including GPS technology to find local services and support, potentially building a community network.
How we built it
The core of the chatbot is based on the natural-language-processing Python module spaCy, which tokenises English input and assigns word vectors to each token. Word vectors are n-dimensional vectors representing the meaning of words, such that similar words are clustered close together. This means that the vectors of individual tokens can be averaged to give the vectors of sentences, and the cosine of the angle between vectors can be used to measure the similarity between two phrases.
The chat bot just has to be able to carry through one of several pre-programmed conversations; this can be represented as a state machine or a graph, which is what the code does. Each state can transition to another one if the user's input text is similar enough to the pre-programmed example responses, like with IBM Watson Assistant; unlike Watson Assistant, though, this is fully offline and so is less likely to be detected by any network-monitoring software an abuser might have set up.
The main part of the app is built in Flutter, which means we only need to re-program some of the confidentiality-ensuring features like "hide on lock" when we port it to iOS; the interface can remain unchanged.
Challenges we ran into
Accomplishments that we're proud of
What's next for Alex News
A facility for documenting and keeping a diary of incidents, situations, feelings and evidence, which could be used by support and legal services to help build evidence for a case.