NLP, Deep learning expert for pattern extraction, custom NER from accident reports
Budget: ₹37,500 – ₹75,000 INR
Hi, Please read the job description carefully and apply if you are capable of doing this job.
I have to extract phrases conveying information like locations, risks, hazards, time, cause from the accident reports. Normal NER techniques are capable of handling person names, locations, dates and times. But this task is to extract information such as risk, hazard, location, time, cause, date from the accident reports. The data is downloaded from internet and it has 1000 rows of useful information, the data is NOT ANNOTATED nor TAGGED. I require your help in using state of art deep learning, machine learning algorithms for completing the task.
1. I will give you a set of seed words under each category. for eg,.
Location = [warehouse, boiler, tank, parking space, workshop],
hazard = [confined space, pressure, work at height, slippery floor, oil floor, low visibility],
risk = [unaware, overlook, caught, struck]
cause = [missing, explosion, fire, smoke, loose, leak etc]
You have to use 'BOOTSTRAPPING" techniques for pattern extraction for extracting all possible words under each category by using the seed words. Also, you have to give the list of patterns that extracted these words. "A useful study with similar methodology is attached for your reference, please read the article before applying".
2. I have already used clustering (T-SNE, K means), topic modelling (LDA, NMF, LSI), word embeddings using BERT, WORD2VEC, FAST TEXT, Glove, So please don't propose any possible method using these techniques, Or if you are keen on applying these techniques for deriving best results please share the proposed idea with me based on your understanding on the subject I will hire you.
3. You have to clean the data, annotate (if required), write the Python code, execute the program, hand over the code and results to me. This code shouldn't be shared on any public platform since I am paying you for this job.
4. You are free to choose any ML techniques other than those mentioned but it should give good results. I can offer high pay if the algorithm or approach is novel and efficient.
The data (please refer to column "summary. new" in the data file) and the useful article is attached, please read the article before applying.
Excellent work deserves additional pay too.
Keep coding, thank you for your attention.
I have to extract phrases conveying information like locations, risks, hazards, time, cause from the accident reports. Normal NER techniques are capable of handling person names, locations, dates and times. But this task is to extract information such as risk, hazard, location, time, cause, date from the accident reports. The data is downloaded from internet and it has 1000 rows of useful information, the data is NOT ANNOTATED nor TAGGED. I require your help in using state of art deep learning, machine learning algorithms for completing the task.
1. I will give you a set of seed words under each category. for eg,.
Location = [warehouse, boiler, tank, parking space, workshop],
hazard = [confined space, pressure, work at height, slippery floor, oil floor, low visibility],
risk = [unaware, overlook, caught, struck]
cause = [missing, explosion, fire, smoke, loose, leak etc]
You have to use 'BOOTSTRAPPING" techniques for pattern extraction for extracting all possible words under each category by using the seed words. Also, you have to give the list of patterns that extracted these words. "A useful study with similar methodology is attached for your reference, please read the article before applying".
2. I have already used clustering (T-SNE, K means), topic modelling (LDA, NMF, LSI), word embeddings using BERT, WORD2VEC, FAST TEXT, Glove, So please don't propose any possible method using these techniques, Or if you are keen on applying these techniques for deriving best results please share the proposed idea with me based on your understanding on the subject I will hire you.
3. You have to clean the data, annotate (if required), write the Python code, execute the program, hand over the code and results to me. This code shouldn't be shared on any public platform since I am paying you for this job.
4. You are free to choose any ML techniques other than those mentioned but it should give good results. I can offer high pay if the algorithm or approach is novel and efficient.
The data (please refer to column "summary. new" in the data file) and the useful article is attached, please read the article before applying.
Excellent work deserves additional pay too.
Keep coding, thank you for your attention.