Social media text analytics of Malayalam-English code-mixed using deep learning

Zigzag conversational patterns of contents in social media are often perceived as noisy or informal text. Unrestricted usage of vocabulary in social media communications complicates the processing of code-mixed text. This paper accentuates two major aspects of code mixed text: Offensive Language Identification and Sentiment Analysis for Malayalam-English code-mixed data set. The proffered framework addresses 3 key points apropos these tasks-dependencies among features created by embedding methods (Word2Vec and FastText), comparative analysis of deep learning algorithms (uni-/bi-directional models, hybrid models, and transformer approaches), relevance of selective translation and transliteration and hyper-parameter optimization-which ensued in F1-Scores (model's accuracy) of 0.76 for Forum for Information Retrieval Evaluation (FIRE) 2020 and 0.99 for European Chapter of the Association for Computational Linguistics (EACL) 2021 data sets. A detailed error analysis was also done to give meaningful insights. The submitted strategy turned in the best results among the benchmarked models dealing with Malayalam-English code-mixed messages and it serves as an important step towards societal good.

Accuracy of conventional cytology: results from a multicentre screening study in India

Objective: We conducted a multi-centre cross-sectional study in India to evaluate the accuracy of conventional cytology to detect high-grade squamous intraepithelial lesions (HSIL).

Setting: Cross-sectional studies in Jaipur, Kolkata, Mumbai and Trivandrum, India, during 1999-2003.

Methods: A common protocol and questionnaire were used to test 22,663 women aged 25-65 years with conventional cytology in five cross-sectional studies. Three thresholds were used to define test positivity: atypical squamous cells of uncertain significance (ASCUS), low-grade squamous intra-epithelial lesion (LSIL), or HSIL. All screened women were investigated with colposcopy, and biopsies were taken when necessary. The reference standard for final disease status was histology or negative colposcopy. Data from the studies were pooled to evaluate the test characteristics for the detection of histologically confirmed HSIL.

Results: The test positivity rates of cytology were 8.8% at ASCUS, 6.2% at LSIL and 1.8% at HSIL thresholds, and 355 women had histologically confirmed HSIL while 74 had invasive cancer. The pooled sensitivity, specificity, positive and negative predictive values at ASCUS threshold were 64.5%, 92.3%, 11.8% and 99.4% respectively. The corresponding values at LSIL threshold were 58.0%, 94.9%, 15.2% and 99.3%, while at the HSIL threshold they were 45.4%, 99.2%, 46.3% and 99.1%. The sensitivity varied between 37.8-81.3% at ASCUS , 28.9-76.9% at LSIL and 24.4-72.3% at HSIL thresholds. A significantly low sensitivity was observed in women aged 25-39 years (p<0.001). The wide variation in sensitivity across study sites persisted even after age standardisation.

Conclusion: The sensitivity of cytology varied widely between the study sites. Findings from our study and other reviews indicate that sustained efforts in improving sampling, preparation and reading of cytological specimens and improvements in clinical judgement are essential to achieve concurrently high sensitivity and specificity.

Completed suicide in adults of rural Kerala: rates and determinants

BACKGROUND

India has witnessed a dramatic increase in suicide rates during the past few decades. The southern state of Kerala has been reporting the highest rates of suicide. Since suicide rates are estimated from death registries, they are likely to be under-reported because the civil registration system is incomplete and suicide deaths are poorly reported.

METHODS

A cohort of 132 000 participants (age 35 years and above) in Thiruvananthapuram (erstwhile Trivandrum) district, Kerala was followed up for mortality from 1996 to 2005, after having filled-in a lifestyle questionnaire at baseline. The cause of death was based on verbal autopsy. Suicide methods were recorded and rates were estimated, and suicide risks were calculated according to several socioeconomic factors.

RESULTS

During the follow up period, a total of 11 608 deaths, of which 385 were suicides (3.3% of total deaths), were registered. The overall suicide rate was 39.3/100 000 person-years among adults 35-90 years of age (men: 78/ 100000; women: 16.5/100000). The predominant methods of suicide were hanging, followed by poisoning and drowning. The suicide determinants were male gender, middle-age (40-60 years), Hindu, alcohol drinkers and secondary education level (< or = 7 years). Neither low socioeconomic level, living alone, nor being a married woman was associated with suicide risk.

CONCLUSION

Suicide rates were consistent with the official rates of Thiruvananthapuram district (37/100 000). However, our study population did not include the 14-34-year-old age-group which represents more than 37% of all suicides and hence it is more likely that the official rates are under-reported. Determinants of suicide were in line with previous studies.

Effectiveness, safety and acceptability of 'see and treat' with cryotherapy by nurses in a cervical screening study in India

We evaluated a ‘see and treat’ procedure involving screening, colposcopy, biopsy and cryotherapy by trained nurses in one-visit in field clinics in a cervical screening study in South India for its acceptability, safety and effectiveness in curing cervical intraepithelial neoplasia (CIN). Women positive on visual inspection with acetic acid (VIA) were advised colposcopy, directed biopsies and cryotherapy if they had colposcopic impression of CIN in one visit by nurses in field clinics supervised by a doctor. Side effects and complications were assessed and cure rates were evaluated with VIA, colposcopy and biopsy if colposcopic abnormalities were suspected. Cure was defined as no clinical or histological evidence of CIN at ⩾6 months from treatment. Of the 2513 women offered ‘see and treat’ procedure, 1879 (74.8%) accepted. Of the 1397 women with histologically proved CIN treated with cryotherapy, 1026 reported for follow-up evaluation. Cure rates were 81.4% (752 out of 924) for women with CIN 1; 71.4% (55 out of 77) for CIN 2 and 68.0% (17 out of 25) for CIN 3. Minor side effects and complications were documented in less than 3% of women. ‘See and treat’ with cryotherapy by nurses under medical supervision is acceptable, safe and effective for cervical cancer prevention in low-resource settings.

Early detection of cervical cancer with visual inspection methods: a summary of completed and on-going studies in India

India is a high-risk country for cervical cancer which accounts a quarter (126,000 new cases, 71,000 deaths around 2,000) of the world burden. The age-standardized incidence rates range from 16-55 per 100,000 women in different regions with particularly high rates in rural areas. Control of cervical cancer by early detection and treatment is a priority of the National Cancer Control Programme of India. There are no organized cytology screening programmes in the country. The technical and financial constraints to organize cytology screening have encouraged the evaluation of visual inspection approaches as potential alternatives to cervical cytology in India. Four types of visual detection approaches for cervical neoplasia are investigated in India: a) naked eye inspection without acetic acid application, widely known as 'downstaging'; b) naked eye inspection after application of 3-5% acetic acid (VIA); c) VIA using magnification devices (VIAM); d) visual inspection after the application of Lugol's iodine (VILI). Downstaging has been shown to be poorly sensitive and specific to detect cervical neoplasia and is no longer considered as a suitable screening test for cervical cancer. VIA, VIAM and VILI are currently being investigated in multicentre cross-sectional studies (without verification bias), in which cytology and HPV testing are also simultaneously evaluated, and the results of these investigations will be available in 2003. These studies will provide valuable information on the average, comparative test performances in detecting high-grade cervical cancer precursors and cancer. Results from pooled analysis of data from two completed studies indicated an approximate sensitivity of 93.4% and specificity of 85.1% for VIA to detect CIN 2 or worse lesions; the corresponding figures for cytology were 72.1% and 91.6%. The efficacy of VIA in reducing incidence of an mortality from cervical cancer and its cost-effectiveness is currently being investigated in two cluster randomized controlled intervention trials in India. One of these studies is a 4-arm trial addressing the comparative efficacy of VIA, cytology and primary screening with HPV DNA testing. This trial will provide valuable information on comparative detection rates of CIN 2-3 lesions by the middle of 2003. The expected outcomes from the Indian studies will contribute valuable information for guiding the development of public health policies on cervical cancer prevention in countries with different levels of socio-economic and health services development and open up new avenues of research. This paper is available too at: http//www.insp.mx/salud/index.html.