![\"\"](\"https:\/\/i0.wp.com\/contributor.insightmediagroup.io\/wp-content\/uploads\/2025\/04\/2025-04-23_23-11-15-ezgif.com-video-to-gif-converter.gif?ssl=1\")
<\/figure>\nTo follow along, you need the following packages:<\/p>\n Using uv<\/a><\/strong>, you can add the dependencies with the following command:<\/p>\n NOTE:<\/strong> When using uv<\/em> with torch you need to specify the index for the package. E.g if you want to use cuda, you need the following in your For these types of projects I always like to start with a quick layout of the UI components. In this case the layout will be quite simple, there\u2019s a textbox with a single line at the top that fills the width and below it the canvas filling the rest of the available space. This will be where we draw the smiley face Using Unfortunately there\u2019s no good out of the box solution for drawing opencv frames on a UI element, so I built my own For our smiley face, we could use different discrete images for sentiment ranges, so for example having three images saved for negative<\/em>, neutral<\/em> and positive<\/em>. However, to get a more fine-grained sentiment visualized, we would need more images and it quickly becomes infeasible and we will not be able to animate transitions between these images. <\/p>\n A better approach is to generate the image of the smiley face procedurally at runtime. To keep it simple, we will only change the background color of the smiley, as well as the curve of its mouth.<\/p>\n First we need to generate a canvas image, on which we can draw the smiley.<\/p>\n Our image should be transparent outside of the smiley face, so we need 4 color channels, the last one will be the alpha channel. Since OpenCV images are represented as We can define some variables for the dimensions and colors of our smiley that will be helpful while drawing.<\/p>\n The background color of the smiley face should be red for negative sentiments and green for positive sentiments. To achieve this with a uniform brightness across the transition, we can use the HSV color space and simply interpolate the hue between 0% and 30%.<\/p>\n We need to make sure to make the color fully opaque by adding a 100% alpha value in fourth channel. Now we can draw our smiley face circle with a border.<\/p>\n So far so good, now we can add the eyes. We calculate an offset from the center to the left and right to place the two eyes symmetrically.<\/p>\n Now on to the challenging part, the mouth. The shape of the mouth will be a parabola scaled appropriately. We can simply multiply the standard parabola In the end the line will be drawn using Et voil\u00e0, we have a procedural smiley face!<\/p>\n To test the function, I wrote a quick test case using To determine how happy or sad our smiley should look like, we first need to analyze the text input and calculate a sentiment<\/em>. This task is called sentiment analysis<\/strong>. We will use a pre-trained transformer model to predict a classification score for the classes NEGATIVE<\/em>, NEUTRAL<\/em> and POSITIVE<\/em>. We can then fuse the confidence scores of these classes to calculate a final sentiment score between -1 and +1.<\/p>\n Using the pipeline from the transformers library, we can define processing pipeline based on a pre-trained model from huggingface.<\/a> Using the To run the sentiment analysis, we can call the pipeline with a string argument. This will return a list of results with a single element, so we need to unpack the first element.<\/p>\n We can define a label mapping, that tells us how each confidence score affects the final sentiment. Then we can aggregate the positivity over all confidence scores.<\/p>\n To test our pipeline, we can wrap it in a class and run some tests using Now the last part that is missing, is simply hooking up the text box to our sentiment pipeline and updating the displayed image with the corresponding smiley face. We can add a And finally the smiley is visualized in the application and changes dynamically with the sentiment of the text input!<\/p>\n For the full implementation and more details, checkout the project repository on GitHub:<\/p>\n https:\/\/github.com\/trflorian\/sentiment-analysis-viz<\/a><\/p>\n All visualizations in this post were created by the author.<\/em><\/p>\n The post Real-Time Interactive Sentiment Analysis in Python<\/a> appeared first on Towards Data Science<\/a>.<\/p>\n<\/div>\nPrerequisites<\/h2>\n
\n
uv add customtkinter opencv-Python<\/a> torch transformers<\/code><\/pre>\n\n
pyproject.toml<\/code>:<\/p>\n[[tool.uv.index]]\nname = \"pytorch-cu118\"\nurl = \"https:\/\/download.pytorch.org\/whl\/cu118\"\nexplicit = true\n\n[tool.uv.sources]\ntorch = [{ index = \"pytorch-cu118\" }]\ntorchvision = [{ index = \"pytorch-cu118\" }]<\/code><\/pre>\n<\/blockquote>\nUI Layout Skeleton<\/h2>\n
![\"\ud83d\ude42\"](\"https:\/\/i0.wp.com\/s.w.org\/images\/core\/emoji\/15.0.3\/72x72\/1f642.png?ssl=1\")
<\/p>\ncustomtkinter<\/code>, we can write the layout as follows:<\/p>\nimport customtkinter\n\nclass App(customtkinter.CTk):\n def __init__(self) -> None:\n super().__init__()\n\n self.title(\"Sentiment Analysis<\/a>\")\n self.geometry(\"800x600\")\n\n self.grid_columnconfigure(0, weight=1)\n self.grid_rowconfigure(0, weight=0)\n self.grid_rowconfigure(1, weight=1)\n\n self.sentiment_text_var = customtkinter.StringVar(master=self, value=\"Love\")\n\n self.textbox = customtkinter.CTkEntry(\n master=self,\n corner_radius=10,\n font=(\"Consolas\", 50),\n justify=\"center\",\n placeholder_text=\"Enter text here...\",\n placeholder_text_color=\"gray\",\n textvariable=self.sentiment_text_var,\n )\n self.textbox.grid(row=0, column=0, padx=20, pady=20, sticky=\"nsew\")\n self.textbox.focus()\n\n self.image_display = CTkImageDisplay(self)\n self.image_display.grid(row=1, column=0, padx=20, pady=20, sticky=\"nsew\")<\/code><\/pre>\nCTkImageDisplay<\/code>. If you want to learn in detail how it works, check out my previous post<\/a>. In short, I use a CTKLabel<\/code> component and decouple the thread that updates the image from the GUI thread using a synchronization queue.<\/p>\n![\"\"](\"https:\/\/i0.wp.com\/contributor.insightmediagroup.io\/wp-content\/uploads\/2025\/05\/image.png?ssl=1\")
<\/figure>\nProcedural Smiley<\/h2>\n
![\"discrete](\"https:\/\/i0.wp.com\/contributor.insightmediagroup.io\/wp-content\/uploads\/2025\/05\/image-1-1024x315.png?resize=1024%2C315&ssl=1\")
<\/figure>\n![\"continuous](\"https:\/\/i0.wp.com\/contributor.insightmediagroup.io\/wp-content\/uploads\/2025\/05\/image-2-1024x369.png?resize=1024%2C369&ssl=1\")
<\/figure>\ndef create_sentiment_image(positivity: float, image_size: tuple[int, int]) -> np.ndarray:\n \"\"\"\n Generates a sentiment image based on the positivity score.\n This draws a smiley with its expression based on the positivity score.\n\n Args:\n positivity: A float representing the positivity score in the range [-1, 1].\n image_size: A tuple representing the size of the image (width, height).\n\n Returns:\n A string representing the path to the generated sentiment image.\n \"\"\"\n width, height = image_size\n frame = np.zeros((height, width, 4), dtype=np.uint8)\n\n # TODO: draw smiley\n\n return frame<\/code><\/pre>\nnumpy<\/code> arrays with unsigned 8-bit integers<\/em>, we create the image using the np.uint8<\/code> data type. Remember that the arrays are stored y-first<\/em>, so the height<\/code> of the image_size<\/code> is passed first to the array creation<\/p>\n color_outline = (80,) * 3 + (255,) # gray\n thickness_outline = min(image_size) \/\/ 30\n center = (width \/\/ 2, height \/\/ 2)\n radius = min(image_size) \/\/ 2 - thickness_outline<\/code><\/pre>\n![\"\"](\"https:\/\/i0.wp.com\/contributor.insightmediagroup.io\/wp-content\/uploads\/2025\/05\/image-16.png?ssl=1\")
<\/figure>\ncolor_bgr = color_hsv_to_bgr(\n hue=(positivity + 1) \/ 6, # positivity [-1,1] -> hue [0,1\/3]\n saturation=0.5,\n value=1,\n)\ncolor_bgra = color_bgr + (255,)<\/code><\/pre>\ncv2.circle(frame, center, radius, color_bgra, -1) # Fill\ncv2.circle(frame, center, radius, color_outline, thickness_outline) # Border<\/code><\/pre>\n![\"\"](\"https:\/\/i0.wp.com\/contributor.insightmediagroup.io\/wp-content\/uploads\/2025\/05\/image-20.png?ssl=1\")
<\/figure>\n# calculate the position of the eyes\neye_radius = radius \/\/ 5\neye_offset_x = radius \/\/ 3\neye_offset_y = radius \/\/ 4\neye_left = (center[0] - eye_offset_x, center[1] - eye_offset_y)\neye_right = (center[0] + eye_offset_x, center[1] - eye_offset_y)\n\ncv2.circle(frame, eye_left, eye_radius, color_outline, -1)\ncv2.circle(frame, eye_right, eye_radius, color_outline, -1)<\/code><\/pre>\n![\"\"](\"https:\/\/i0.wp.com\/contributor.insightmediagroup.io\/wp-content\/uploads\/2025\/05\/image-22.png?ssl=1\")
<\/figure>\ny=x\u00b2<\/code> with the positivity score.<\/p>\n![\"\"](\"https:\/\/i0.wp.com\/contributor.insightmediagroup.io\/wp-content\/uploads\/2025\/05\/image-24-1024x659.png?resize=1024%2C659&ssl=1\")
<\/figure>\ncv2.polylines<\/code>, which needs xy coordinate pairs. Using np.linspace<\/code> we generate 100 points on the x-axis and the polyval<\/code> function to calculate the according y values of the polygon.<\/p>\n# mouth parameters\nmouth_wdith = radius \/\/ 2\nmouth_height = radius \/\/ 3\nmouth_offset_y = radius \/\/ 3\nmouth_center_y = center[1] + mouth_offset_y + positivity * mouth_height \/\/ 2\nmouth_left = (center[0] - mouth_wdith, center[1] + mouth_offset_y)\nmouth_right = (center[0] + mouth_wdith, center[1] + mouth_offset_y)\n\n# calculate points of polynomial for the mouth\nply_points_t = np.linspace(-1, 1, 100)\nply_points_y = np.polyval([positivity, 0, 0], ply_points_t) # y=positivity*x\u00b2\n\nply_points = np.array(\n [\n (\n mouth_left[0] + i * (mouth_right[0] - mouth_left[0]) \/ 100,\n mouth_center_y - ply_points_y[i] * mouth_height,\n )\n for i in range(len(ply_points_y))\n ],\n dtype=np.int32,\n)\n\n# draw the mouth\ncv2.polylines(\n frame,\n [ply_points],\n isClosed=False,\n color=color_outline,\n thickness=int(thickness_outline * 1.5),\n)<\/code><\/pre>\n![\"\"](\"https:\/\/i0.wp.com\/contributor.insightmediagroup.io\/wp-content\/uploads\/2025\/05\/image-25.png?ssl=1\")
<\/figure>\npytest<\/code> that saves the smiley faces with different sentiment scores:<\/p>\nfrom pathlib import Path\n\nimport cv2\nimport numpy as np\nimport pytest\n\nfrom sentiment_analysis.utils import create_sentiment_image\n\nIMAGE_SIZE = (512, 512)\n\n\n@pytest.mark.parametrize(\n \"positivity\",\n np.linspace(-1, 1, 5),\n)\ndef test_sentiments(visual_output_path: Path, positivity: float) -> None:\n \"\"\"\n Test the smiley face generation.\n \"\"\"\n image = create_sentiment_image(positivity, IMAGE_SIZE)\n\n assert image.shape == (IMAGE_SIZE[1], IMAGE_SIZE[0], 4)\n\n # assert center pixel is opaque\n assert image[IMAGE_SIZE[1] \/\/ 2, IMAGE_SIZE[0] \/\/ 2, 3] == 255\n\n # save the image for visual inspection\n positivity_num_0_100 = int((positivity + 1) * 50)\n image_fn = f\"smiley_{positivity_num_0_100}.png\"\n cv2.imwrite(str(visual_output_path \/ image_fn), image)\n<\/code><\/pre>\n![\"\"](\"https:\/\/i0.wp.com\/contributor.insightmediagroup.io\/wp-content\/uploads\/2025\/05\/image-26-1024x252.png?resize=1024%2C252&ssl=1\")
<\/figure>\nSentiment Analysis<\/h2>\n
top_k<\/code> parameter, we can specify how many classification results should be returned. Since we want all three classes, we set it to 3.<\/p>\nfrom transformers import pipeline\n\nmodel_name = \"cardiffnlp\/twitter-roberta-base-sentiment\"\n\nsentiment_pipeline = pipeline(\n task=\"sentiment-analysis\",\n model=model_name,\n top_k=3,\n)<\/code><\/pre>\nresults = self.sentiment_pipeline(text)\n\n# [\n# [\n# {\"label\": \"LABEL_2\", \"score\": 0.5925878286361694},\n# {\"label\": \"LABEL_1\", \"score\": 0.3553399443626404},\n# {\"label\": \"LABEL_0\", \"score\": 0.05207228660583496},\n# ]\n# ]\n\nfor label_score_dict in results[0]:\n label: str = label_score_dict[\"label\"]\n score: float = label_score_dict[\"score\"]<\/code><\/pre>\nlabel_mapping = {\"LABEL_0\": -1, \"LABEL_1\": 0, \"LABEL_2\": 1}\n\npositivity = 0.0\nfor label_score_dict in results[0]:\n label: str = label_score_dict[\"label\"]\n score: float = label_score_dict[\"score\"]\n\n if label in label_mapping:\n positivity += label_mapping[label] * score<\/code><\/pre>\npytest<\/code>. We verify that sentences with a positive sentiment have a score greater than zero and vice versa sentences with a negative sentiment should have a score below zero.<\/p>\nimport pytest\n\nfrom sentiment_analysis.sentiment_pipeline import SentimentAnalysisPipeline\n\n\n@pytest.fixture\ndef sentiment_pipeline() -> SentimentAnalysisPipeline:\n \"\"\"\n Fixture to create a SentimentAnalysisPipeline instance.\n \"\"\"\n return SentimentAnalysisPipeline(\n model_name=\"cardiffnlp\/twitter-roberta-base-sentiment\",\n label_mapping={\"LABEL_0\": -1.0, \"LABEL_1\": 0.0, \"LABEL_2\": 1.0},\n )\n\n\n@pytest.mark.parametrize(\n \"text_input\",\n [\n \"I love this!\",\n \"This is awesome!\",\n \"I am so happy!\",\n \"This is the best day ever!\",\n \"I am thrilled with the results!\",\n ],\n)\ndef test_sentiment_analysis_pipeline_positive(\n sentiment_pipeline: SentimentAnalysisPipeline, text_input: str\n) -> None:\n \"\"\"\n Test the sentiment analysis pipeline with a positive input.\n \"\"\"\n assert (\n sentiment_pipeline.run(text_input) > 0.0\n ), \"Expected positive sentiment score.\"\n\n\n@pytest.mark.parametrize(\n \"text_input\",\n [\n \"I hate this!\",\n \"This is terrible!\",\n \"I am so sad!\",\n \"This is the worst day ever!\",\n \"I am disappointed with the results!\",\n ],\n)\ndef test_sentiment_analysis_pipeline_negative(\n sentiment_pipeline: SentimentAnalysisPipeline, text_input: str\n) -> None:\n \"\"\"\n Test the sentiment analysis pipeline with a negative input.\n \"\"\"\n assert (\n sentiment_pipeline.run(text_input) < 0.0\n ), \"Expected negative sentiment score.\"\n<\/code><\/pre>\nIntegration<\/h2>\n
trace<\/code> to the text variable, which will run the sentiment pipeline in a new thread managed by a thread pool, to prevent the UI from freezing while the pipeline is running.<\/p>\nclass App(customtkinter.CTk):\n def __init__(self, sentiment_analysis_pipeline: SentimentAnalysisPipeline) -> None:\n super().__init__()\n self.sentiment_analysis_pipeline = sentiment_analysis_pipeline\n\n ...\n\n self.sentiment_image = None\n\n self.sentiment_text_var = customtkinter.StringVar(master=self, value=\"Love\")\n self.sentiment_text_var.trace_add(\"write\", lambda *_: self.on_sentiment_text_changed())\n\n ...\n\n self.update_sentiment_pool = ThreadPool(processes=1)\n\n self.on_sentiment_text_changed()\n\n def on_sentiment_text_changed(self) -> None:\n \"\"\"\n Callback function to handle text changes in the textbox.\n \"\"\"\n new_text = self.sentiment_text_var.get()\n\n self.update_sentiment_pool.apply_async(\n self._update_sentiment,\n (new_text,),\n )\n\n def _update_sentiment(self, new_text: str) -> None:\n \"\"\"\n Update the sentiment image based on the new text input.\n This function is run in a separate process to avoid blocking the main thread.\n\n Args:\n new_text: The new text input from the user.\n \"\"\"\n positivity = self.sentiment_analysis_pipeline.run(new_text)\n\n self.sentiment_image = create_sentiment_image(\n positivity,\n self.image_display.display_size,\n )\n\n self.image_display.update_frame(self.sentiment_image)\n\n\ndef main() -> None:\n # Initialize the sentiment analysis pipeline\n sentiment_analysis = SentimentAnalysisPipeline(\n model_name=\"cardiffnlp\/twitter-roberta-base-sentiment\",\n label_mapping={\"LABEL_0\": -1, \"LABEL_1\": 0, \"LABEL_2\": 1},\n )\n\n app = App(sentiment_analysis)\n app.mainloop()\n<\/code><\/pre>\n![\"\"](\"https:\/\/i0.wp.com\/contributor.insightmediagroup.io\/wp-content\/uploads\/2025\/05\/2025-05-04_17-39-55-ezgif.com-video-to-gif-converter.gif?ssl=1\")
<\/figure>\n
\n
\n
\n