The diagnosis of tuberculous pleural effusion (TBPE) is frequently problematic. Several markers of TBPE in pleural fluid have been evaluated, with different results.Pleural effusions from 96 patients were classified on the basis of definitive diagnosis as tuberculous (n = 39), neoplastic (n = 42) or parapneumonic (n = 15). Adenosine deaminase (ADA), ADA isoform ADA-2, interferon (IFN)-γ, CD3⁺/DR⁺T-lymphocytes and interleukin (IL)-12 p40 were determined in all 96 effusions. The efficiency of IL-12 p40 for diagnosis of TBPEs was evaluated, in comparison with those of the other parameters, by comparing the areas under their receiver operating characteristics.With the threshold value of 550 pg·mL⁻¹, IL-12 p40 had a sensitivity of 92.3% (36 out of 39) and specificity of 70.2% (17 false positives). The misclassification rate of IL-12 p40 was significantly greater than those of ADA-2 and ADA. Among TBPEs, ADA correlated significantly with ADA-2, and IFN-γ with ADA and IL-12 p40.Although tuberculous pleural effusions show values of interleukin-12 p40 that are significantly higher than neoplastic and parapneumonic fluids, this parameter is less efficient than adenosine deaminase, adenosine deaminase isoform 2 and interferon-γ. Its routine determination is, accordingly, not justified.

Nowadays, there are many types of viral foods and consumers expect to be able to quickly find foods that meet their own tastes. Traditional recommendation systems make recommendations based on the popularity of viral foods or user ratings. However, because of the different sentimental levels of users, deviations occur and it is difficult to meet the user’s specific needs. Based on the characteristics of viral food, this paper constructs a hybrid recommendation approach based on viral food reviews and label attribute data. A user-based recommendation approach is combined with a content-based recommendation approach in a weighted combination. Compared with the traditional recommendation approaches, it is found that the hybrid recommendation approach performs more accurately in identifying the sentiments of user evaluations, and takes into account the similarities between users and foods. We can conclude that the proposed hybrid recommendation approach combined with the sentimental value of food reviews provides novel insights into improving the existing recommendation system used by e-commerce platforms.