This article introduces NLP, it offers many business use cases. He speeds up CV analysis and recruitment, allows theanalysis of customer reviews For improve marketing, facilitates customer service via chatbots, helps with medical research and the fight against COVID-19, improves workplace safety, ensures the protection of personal data And detects fraud.
The use of automatic language processing (NLP or NLP in English) during the recruitment phase helps speed up the analysis of CVs and the search for candidates by identifying the profiles that best match the position to be filled.
First, we identify the relevant keywords after removing the bias and gender present in the job description and we generate a corpus of synonyms.
Then, all the candidates' CVs are compared with this reference keyword base previously defined in order to identify those which most correspond to the desired profile. This first processing optimizes the time spent analyzing CVs, a highly time-consuming task for recruiters who receive hundreds or even thousands of CVs for each job offer.
Customer reviews are present everywhere on the internet and in considerable numbers: online sales sites, forums, social networks, etc., which makes their valuation very time-consuming if done manually. NLP makes it possible to analyze all of these comments, identify feelings relating to a brand's product or service and adapt it to best meet customer needs.
NLP also makes it possible to highlight themes addressed in these opinions. Coupled with sentiment analysis, this can highlight the strong points and weak points of a brand for example (for a restaurant, this could be positively the theme linked to staff and hospitality for example ). Aqsone has developed a review sentiment analysis solution for Toulouse restaurants. This solution allows restaurateurs to improve their service based on customer feedback and to know their strengths.
Natural language processing also helps identify new audiences potentially interested in certain products. Indeed, we can find key information by analyzing textual data that has been extracted from different blogs, websites, or publications on social networks. This allows brands to effectively expand their communication channels and identify the most suitable sites or social networks to place their advertisements and reach their customers.
Chatbots, the vast majority of which are based on NLP technologies, can provide fast and efficient customer service by answering routine questions and handling simple requests at any time of the day.
Customer satisfaction is even better, and makes it possible to optimize the size of teams dedicated to customer service by only assigning complex tasks to humans.
The capabilities of these chatbots are evolving enormously. As evidenced by this video from Google, which presented Google Duplex, a voice chatbot capable of having an advanced conversation with a human:
Although the most popular AI technology in the health field is computer vision (automatic detection of tumors on medical imaging for example), NLP is not far behind. Indeed, thanks to this technology, large-scale automated targeting of patients suffering from this or that pathology is possible. For example, by cross-referencing all historical medical data of patients such as computerized files, hospitalization reports and other medical records, researchers from Yale University in the United States retrospectively identified patients with history of carotid stenosis (Source: “Identification of patients with carotid stenosis using natural language processing”).
Furthermore, in these times of health crisis due to COVID-19, NLP has been of great importance in identifying how the virus works and developing a treatment. Indeed, scientists have used this technology to study protein sequences and determine the genetic backbone of the virus. This operation was possible by considering a protein, which is a sequence of amino acids, as a language where amino acids are the alphabet (Source: “Natural Language Processing in the fight against COVID-19”).
Work safety is an essential aspect on a production or assembly line. Thus, many companies are launching projects with the ambition of using as much available data as possible (HR data, production data, prevention and safety data, location data, etc.) to be able to better understand accidents on a site and identify potential problems. action levers that will reduce the risk of accidents as much as possible.
At Aqsone, we have applied NLP to accident reports. This made it possible to highlight the recurring appearance of certain terms linked to equipment, tools or environmental contexts (slippery floors for example).
The results of this analysis led to the implementation of new safety actions by the profession which made it possible to positively impact accident trends at our client.
Protection of personal data
The protection of personal data has become a real issue in order to be in full compliance with the GDPR (General Data Protection Regulation). The objective is therefore to help data protection managers by creating document analysis tools capable of identifying the presence of personal data within very short time frames and compatible with legal requirements. NLP makes it possible to browse a large quantity of files and highlight any element that may be related to personal data (RIB, medical examinations, etc.). Once sensitive documents have been identified, the department responsible for data protection can take the necessary actions to ensure their protection. Aqsone has developed a solution that allows personal data to be identified, which allowed a client to qualify the nature of the documents in order to protect them in the event of a cyber attack.
Internal fraud costs French companies 5 % of their turnover. There is therefore a real challenge in putting in place the necessary actions to fight against it. Aqsone worked for these clients on expense report fraud detection. Thanks to NLP it was possible to identify the most useful markers for fraud detection, the following information was extracted from the tickets: the total amount, the date, the location, the presence or absence of alcohol, and the presence or not of a supermarket name. These different characteristics were obtained using different NLP techniques.