Speeches have always held a great amount of importance to garner the attention of the crowd. Since ancient times, political figures have used strong speeches to show their opinions in the minds of the public. At this time, political figures depend on speech writers for powerful speeches. While this has been the most successful way of delivering speeches so far, technology has now made it easier for creating speeches with the help of the speech generator.
With the help of the speech generator, political parties will be able to generate political speeches. The system will be built in a way that the party can choose whether the speech should be in support of an opinion or if it should be opposing an opinion.
A sequence of words is generated based on probabilities. These probabilities are obtained from two underlying models. One is a language model which ensures that the speech is grammatically correct. Then there is the topic model which ensures that the speech is fluid and has textual consistency.
The two models were trained on the Convote dataset. The transcriptions contained in it are from the US Congressional floor debates. The Political Speech Generator relies on a combination of NLP methods which include n-grams, recurrent neural networks, Justeson and Katz POS tag filter as well as latent Dirichlet allocation.
To evaluate the quality of the speeches, a manual as well as automated approach has been adopted. To test the quality of the speeches, a few experimental evaluations were conducted. These evaluations have shown that the political speech generator is capable of producing very high quality speeches. The grammatical correctness of the speeches were good and they also showed great sentence transitions.
With the political speech generator, political figures will be able to come up with powerful speeches without having to rely on speech writers.